RethinkDB: why we failed (2017)
defmacro.org
defmacro.org
People that love database technology — that would be me — tend to start database companies. It is very difficult to sell a database. It is much, much easier to sell a compelling solution to a somewhat boring but very valuable business problem that just happens to require an amazing database capability behind it. That’s your moat, the customer doesn’t actually care that there is an amazing database engine behind it but it makes it difficult for competitors to replicate.
Filed under “lessons I learned the hard way”.
Infrastructure tech makes for a great case that helps everyone while giving nobody a competitive advantage, that's why companies love doing or sponsoring it.
On the other hand, open source tech in infrastructure helps get rid of vendor lock-in as in the worst case, you could still look inside the source code and patch things — also fancy FOSS tech makes for good community building efforts and shiny job ads, both attracting good devs.
Both effects together explain to me why open source makes for such a fierce competitor in any kind of paid infra tech, it's hard to build the momentum to outrun them.
And open core? Pretty much dead by now since as soon as the cloud providers see there's traction around any technology, they will build a hosted version, either directly or in a protocol compatible fashion.
Techies really dig this stuff (I do too), but as a business model, there aren't many worse ones in 2021.
A question that lacks a satisfactory answer in open source data infrastructure is this: who is going to pay for the initial architecture to be state-of-the-art? Architecture is forever, if you start with a naive design it will put the platform at a long-term disadvantage. Most open source infrastructure projects are started by well-meaning people that have no idea what a state-of-the-art design even looks like. At the periphery you might be able to burn a giant pile of VC money to make it happen but that is not sustainable without a clear path to profitable exit. The set of people with the critical expertise and the set of people building open source data infrastructure are nearly disjoint.
Today, the money is in state-of-the-art bespoke data infrastructure. If the tradeoff is "open source" and "10-100x better on infrastructure KPIs", companies happily pay millions for the latter. People that know how to design this infrastructure are extremely well paid, 7-figures is common and demand is high. Yes, they could take a year off work and create an open source infrastructure project but this doesn't seem to happen and it is easy to understand why, as there are few incentives and many disincentives. This is roughly my area of business, I see the dynamics from the inside and the trend is not favoring open source.
If all of the cutting edge infrastructure development work is happening in closed source, whether cloud or state-of-the-art bespoke, it doesn't bode well for the long-term relevance of open source.
SQL Server, Oracle just led to AWS DynamoDB/Aurora. You don't have any source code for either. Open-source projects growing in popularity doesn't override the greater trend that companies just want solutions, whether it's open or not.
AWS had a problem or “Hosted Super Open Source Database” had a problem.
There is a reason why there is an old adage: “No one got fired for going with (IBM/Microsoft/AWS)”
And then in the old on-premise world, you sell them both the App and require the use of your database which is an additional sale.
This strategy doesn’t work in a Cloud world though. Since customers are no longer buying the individual components (like the database) but are instead just leasing the finished good. By definition, they don’t care what’s under the hood - unless you’re selling to another software company who’s using your database to make their finished good.
That provided enough funding, users and support to move into applications. In fact even if they didn't expand into other areas, that support could have kept them as a profitable DB company for decades.
Your comment reminds me of April Dunford's post on how she helped position a database product: https://www.thefxck.com/interviews/product-positioning-april...
So the internal tech is just a mean to make that possible. I think if a db engine provide the equivalent of the auto-admin of Django it will sell itself easily :)
However this is also hard because DBs engines touch several complex things (storage, concurrency, compilers, interpreters, etc) and is hard to find people for this plus the funding.
One thing these database companies have going for themselves, is that once a customer company’s data and code is locked in to a particular database, then it becomes very very hard to get out. Not impossible, but just very difficult.
However, at least three issues make a database migration non-trivial:
1. Custom data types that have to be replicated somehow in the new database; this usually involves figuring out how to parse the output representation in a sensible way.
2. User-defined triggers and functions, or custom database features, that may not be available on the new database.
3. There is usually a lot of infrastructure built on top of a database that can be hard to switch over, like if you've used roles substantially in Postgres for access control. Not to mention any business intelligence tool built on it that tends to have a lot of hand-rolled SQL. For example, Periscope can be database-agnostic, but any queries you write might be customized for one particular database vendor.
So in my view, it's not really data concerns (except case 1); there are a lot of operational concerns that make database migrations hard.
Well, things weren't this easy. It was Oracle alright, so all the queries, triggers and views still worked... except the performance characteristics were completely different, because Exadata does all kinds of interesting things to extract more performance. So all the hints, manual query plans? just downright wrong, and often slower than before. In the end, easily a third of the queries needed rewrites, plus all the internal tuning changes that the very large DBA team did to make queries perform well. The whole effort cost many thousands of man months, on top of the oracle hardware, and the never cheap oracle license.
So even a migration from Oracle to a different flavor of Oracle can end up freezing a department with hundreds of tech workers for a year, and this was considered to be the cheapest choice! Imagine how much fun this would be if it was a large lift that, say, went from a RDBMS to NoSQL.
But then the original licensing agreement we had negotiated was coming to an end, and the new deal was going to be massively more expensive.
So we used this as an opportunity to rearchitect our monolith into services built on other open source data stores, that also fixed a lot of the architectural issues and technical debt we had acquired.
So to answer your question, yes it’s possible to move off of a database, if the financial incentives are great enough.
Some context: this is for expansive legacy application databases that have many years of tenure or large data warehouse applications. If you've ever worked in the enterprise data space, you'll be tripping over this sort of workload left and right.
"You don't need us to install any database software?"
"nah we have our own integrated data persistence mechanism".
"Cool. So about those firewall rules..."
That is about the extent to which our customers care. They are more interested in what actually shows up on their display and how stable the solution is than any particular technology choices we decided to make on their behalf.
No one wants to be responsible for picking a database engine. Pick it for your customers. That's part of the engineering.
"What is it going to take for you to get the message that customers don’t want the things that architecture astronauts just love to build. The people? They love twitter. And flickr and delicious and picasa and tripit and ebay and a million other fun things, which they do want, and this so called synchronization problem is just not an actual problem, it’s a fun programming exercise that you’re doing because it’s just hard enough to be interesting but not so hard that you can’t figure it out."
https://www.joelonsoftware.com/2008/05/01/architecture-astro...
"As computing technology develops it becomes more efficient to take centralised computing resources and distribute them closer to the user. As network technology develops it becomes more efficient to centralise them again. Further advances redistribute and yet further advances recentralise. This pattern has been noticed several times in the history of computation."
And the most fascinating thing was that it was from decades ago, perhaps even 1960! I've never been able to find the quote again. Does anyone recognise it? Perhaps I imagined it.
Also, I wonder if commodity cloud offerings such as AWS will change this?
https://youtu.be/O9upVbGSBFo?t=340 (05:40 - 10:50)
Analogous to it, the nosql DB market is the exact wrong market to enter. Large companies willing to pay big bucks for enterprise features will pay established vendors for the stodgy but battle tested stuff like Oracle or DB2. The hipster startup market pays nobody for anything as there is a myriad of free choices in every common flavour and the few that will pay will mostly do so to purchase managed hosting. And that’s only if their PoC built on your new and untrusted database ever makes it to production. And then it’s probably your cheapest tier. Don’t be selling to paupers!
But AWS is killing this model by just offering popular open source products as a service themselves, cutting out the original developments.
They are absolutely still a thing in every major city.
Only if the owner was trying to sell Coffee in the first place, which allegedly isn't Starbucks' main business: https://archive.is/9TZej
This is key, and plays out over and over again in different forms. There are no points for difficulty, only supply and demand. PG puts this well [1]:
> That's the essence of a startup: having brilliant people do work that's beneath them. Big companies try to hire the right person for the job. Startups win because they don't—because they take people so smart that they would in a big company be doing "research," and set them to work instead on problems of the most immediate and mundane sort. Think Einstein designing refrigerators.
In the era of COVID, it seems like every developer with some extra free time has decided that they want to build their own PKM/note-taking/etc app, so competition has scaled up massively in the last year.
Still none with dark mode, native apps, markdown.
I suspect by asking that question any reader starts immediately scratching their head thinking up what problem/solution that could possibly be. Don’t. That confuses product for business.
So yeah, that is speaking of demand. Economically-relevant demand.
RethinkDB was done really, really well. It is one of the very few distributed databases that went through Jepsen relatively unscathed and delivered on promises made. Development was done in the public, questions were asked through StackOverflow. You interacted with competent, skillful and experienced developers.
And yet MongoDB was the latest fashion fad, end even though it did NOT deliver on the promises, it was the hot-database-du-jour that kids used.
The article is mostly about the business model, and I also always thought they would have a hard time making money on the database. But I think they would have had a much better shot if it became more popular. It didn't, which accelerated the company's demise.
EDIT: I just realized that I wrote about RethinkDB in the past tense, even though it very much exists as I write it. In fact I run my business on it. But because it fell out of favor, when it was open-sourced, it failed to pick up momentum, and it now seems unlikely that it will.
Which is why I'm working on switching to FoundationDB, and I'm slightly worried that it will suffer the same fate: it is excellent technically (the best transactional guarantees in a distributed database you can get), but difficult to understand and not very user-friendly. It's not the "node.js database for everyone". The only reason I'm considering it is because Apple uses and develops it, which gives me hope for longer-term maintenance.
Going back to fashions — you can have a product which excels technically, but if it's out of fashion, it might as well not exist.
I think we would all be better off if we stopped trying to always pick The One True Database, The One True Programming Language, etc — and instead accepted that there might be multiple tools, each specialized for certain kinds of tasks.
The segment of the market that chases fashions and fads is dominated by fashions and fads. Most development is just using something old and boring like Postgres or MySQL or MSSQL or Oracle.
The problem for RethinkDB is that when you take out the fashions and fads segment and the old and boring segment, there's not much left.
As a former RethinkDB user (we have migrated to Postgres) I actually don't miss it as much as I would - JSONB in PG does what we need, and the real-time features of RethinkDB never really delivered because of various performance issues in the database itself.
For me, JSON wasn't the important part. I wanted to have a distributed database, where a single server could disappear completely at any time and my customers would not lose any data.
There are remarkably few solutions that deliver this, though many claim to. Even fewer have been verified with Jepsen. And for all the hoopla about distributed computing using Kubernetes, I find it puzzling that so little emphasis is placed on data integrity.
Changefeeds have worked great for me as well, and I'm still trying to work out what to replace them with.
That's not to say that they would have succeeded if they'd gone after that though; instead we might have a blog post about how they correctly predicted where the market would go and simply ran out of money before the market got there.
- Changefeeds were so useful and still to this day not really matched in quality.
- Official client libraries were very high quality. I mainly used Node.js.
- Performance (reads and writes) was impressive.
- Setting up a highly available cluster with sharding was incredibly easy with a few clicks and types in the aforementioned web U/I.
- Their web U/I was the best to ship with a database at the time.
The best technology and product doesn't always win.
To the point that retrieving data based on a secondary index (of like 10000 documents) took like 5s.
Then inserting the same documents took a significant (don’t remember the exact magnitude) amount of time as well.
By comparison mysql does the same stuff in 50ms (or less).
Everything around RethinkDB was fantastic, but I just couldn’t justify that since the basic operation was so terrible.
MongoDB is a weird one because what they had didn't even meet the most basic requirements of "a database", that you could store something in it and reliably get it back out. But they somehow converted their joke of a product into enough money to buy WiredTiger, rebrand its product as theirs, and profit. It's the ultimate expression of "fake it 'til you make it" and they did it by giving away stickers at conferences!
The best technology and product doesn't always win.
Once you realise what we do is not engineering, it's fashion, it will all make sense.
It goes to show that HackerNews consensus != reality.
I do wonder where/how Mongo makes their money - if anybody has some insight, please share.
Edit: Mongo is a special kind of lock-in: your code ends up so much more tightly-coupled, and hard to extricate, because of the Mongo pattern of making you do the work to ensure consistency and correctness. I can’t imagine it was planned, but it’s highly effective: for many companies it’s just too much work, too expensive, and the benefits are too opaque (“but the application works now, and you’re telling me you want to switch databases and it’ll take how long?!”) for others to buy into.
I’d call it clever, but I think it’s a profitable accident that’s a braking force on the progress of many teams.
And it is it not just me and my friends -- HSBC recently announced they are moving 65 of its relational databases to mongodb. I am not sure that is wise, but I think it furthers my view that data-loss is not a concern in many contexts.
The data-loss thing was most certainly real, but has not been since wiredtiger if one manages it the right way or uses it as a service.
I would not say that I would say that HN != reality, maybe !== reality :). Neither side of the debate is wholly right, and neither is wholly wrong.
The reason I use it is that for many projects, all you need is a relatively fast JSON document store, and mongo is so developer-friendly such that speed-to-prototype is so much shorter than anything else I have used in 25 years of building database-backed applications. Rethink was the only product that I would have used in its place and I was extremely sad to see it go. I agree with the commenter elsewhere in the belief that it would have thrived had it survived. I was using it for my personal/learning projects and was ready to switch everywhere.
We have stuck with Mongo in production for a few products, one relatively large, and I think if you have separated your concerns well-enough, it is okay developing in Mongo while leaving yourself a path to postgresql or something else. You need to prove, first, that the thing you are building is useful, and Mongo will let you get to that really fast.
I believe what is happening here is that the happy folks are quiet, and the aggrieved folks' experiences are real but it likely was a bad fit for their particular situation. For the subset of situations where we use it, it has allowed my teams quickly in our experimental work.
All of that said, I work in an innovation space where we throw a lot of things at the wall to see what sticks, and our time-to-prototype is often a few months at most. For us, it has been a godsend, but our situation is rather unique.
-h
And back when I started with it, JSONb did not exist. I would say JSONB was also a godsend though as our roadmaps now have a clear path to something we may prefer later. People up the chain are often terrified when they hear the word "mongo" so the existence of JSONB makes it easier to swap out your layer that handles database interactions.
But the speed advantage really lies in the early-stage prototype development. If I have to do it in postgres, the amount of code necessary up front, when inevitable schema changes happen, etc. makes mongo preferable until things settle down.
But again, it all depends on how well you know the thing up front. If you can anticipate a lot of iteration on the data model as you learn how people use your product, mongo is handy early. If I am given a spec for something where all of the things are known and foreseen, I am going with postgresql.
I assume that the hardest thing to do is making something people want to use. If I can achieve that with whatever I use, I have some good (and fun) problems on my hands. And it is all about structuring the work so that I am not tied to any database product. I am no longer working on startups with little-to-no funding, so right now I assume AWS or Azure will manage a database better than I can. And I assume new things are always coming.
> Querying is in pure JS, easy to do stuff programmatically.
You mean you do your logic at the application layer? Like if you need X objects and Y objects where Y have an identifier to the X objects they're related to you build them in JS after retrieving them?
What is the difference from getting a json object and doing the "querying" in pure js?
And isn't it easier "programatically" to just join the data at the source?
> But the speed advantage really lies in the early-stage prototype development. If I have to do it in postgres, the amount of code necessary up front, when inevitable schema changes happen, etc. makes mongo preferable until things settle down.
But again, what is the difference of having a record with a column with json, that you just do rodeo dump?
(I think pg even allows you to do all sorts of indexing on inner fields of json if I'm not mistaken?)
If I want a particular array item nested a few levels in a JSON, the query as JSON object instead of query as an SQL is just a bit easier to put together (which is what i mean by pure JS). Sometimes, putting complex queries together takes fewer lines of code. Neither is really better or worse.
But I want the database doing the work and want to transform as little as possible on the application layer. Sometimes it is inevitable, but sub-optimal, for sure.
In terms of anything where you need to do a lot of joins when querying -- NoSQL is not good and I would not advise Mongo at all. It depends on what you are building. A lot of applications (and even games) now just pass JSONs around. If I show a user profile, I get that json. If, on the profile, I need to show where she lies in a leaderboard, that react component hits a separate API call to get the data needed for it, etc.
But no difference really. And yes, if the JSON is just a column in postgres, even the early stuff is nearly as easy as it is in postgresql.
If I were looking at a DB for a simple key/value store, I would not look at Mongo. Others here likely know better, but I pick up Redis (out of habit, but also it is awesome even if overkill in many scenarios). Mongo would do it but I would not do so except temporarily. I tend to throw Redis in my stacks early in development as I tend to need it later.
BTW by no means do I love Mongo. For example, Mongo is awful when it comes to things simiilar to joining in SQL land. You end up in these hellscapes where there is no choice but to store a ton of redundant data. Is it the end of the world? No, storage is cheap. But is it good? No.
And I am old so I do not find the value in arguing about tools. It is like me hating a hammer. I am interested in what is better and what folks here think about things coming down the pipeline, or whether Arango is worth looking at, etc. But I do appreciate your sort of inquiry as I have learned from it -- for example I did not know you can index inner fields of JSONs. I really need to revisit is as recent versions of PG seem to get better and better.
Context matters a ton. If you are going right to production with something, that is not really where I operate. But in general, as many have said, if more than one thing will do it, the best tool is the one you know best. You are likely to implement it efficiently, etc.
I agree with you on redis-like utilities being better for pure KV stores. If you resort to that sort of thing regularly and want to give another look into postgres you might want to give Elixir a try. It has first class support for postgres with Ecto, and you can use redis-like stores from the language itself, skipping serialization (a note is that, one of the valuable things in redis usually comes from being in a separate instance, many times managed automatically, and so almost being a permanent cache - to have exactly the same properties you would need to dig a bit deeper in Elixir, but still pretty doable with the language building blocks).
The very RAD reasons you note regarding Mongo also encourage Mongo Cancer (tm).
What is Mongo Cancer you ask? It is the phenomena where Mongo use is manifest from API to home-grown "relational logic" in business tier. This simply doesn't happen with any other NoSQL DB afaik, but Mongo Cancer uniquely marches right up to the user's face in the API.
> I think if you have separated your concerns well-enough, it is okay developing in Mongo while leaving yourself a path to postgresql or something else.
I'll hold you to that "I think" and assume you have never actually done this.
Because Mongo Cancer patients typically love the RAD aspect of grabing JSON hot off the http stack and shoving it in the "database", touching every API in the way. The last M.C. project I saw required effectively a rebuild to actually swap the misused Mongo.
tldr; Mongo Cancer is very easy to catch but requires very expensive surgery to cure.
I do not think the "cancer" metaphor is valuable here. What you describe are bad ideas that I agree are bad ideas but really one has to know the tool they are using and what it is bad at.
I think the thing you describe can happen but that is just poor planning more than anything. Why does the database have anything to do with the design of your API layer? That sounds bonkers. And I would not blame the underlying database tech. I would blame the implementer.
But really, please consider that many of us have, or are as we speak, losing loved ones to a deadly sickness which you have compared to a poor database implementation.
> But really, please consider that many of us have, or are as we speak, losing loved ones to a deadly sickness which you have compared to a poor database implementation.
Sorry for your loss. Use of the word for non-medical concerns is common. Mongo Metastasis works too.
Again disagree regarding "poor database implementation". Trivializing the "database implementation" is precisely Mongo's value proposition and the very reason they got their market share.
I would even say the strengths that I have pointed out are precisely what makes it easier to take shortcuts. It gets out of your way, makes collections if they do not exist, allows for different document types in the same collections. It has no guard rails whatsoever.
I think we are in agreement -- one needs to be very careful using this. You can use it and put yourself in a spot where you have to throw out your entire codebase.
Well, me saying "can happen" does not mean it does not happen or that people have not seen it. It is saying the opposite. My point is it does not have to or that it does not always happen.
Developers take shortcuts, often because they do not know better, and you are right that some products make it easier to go down these roads.
I still do not think you want to use medical diseases as metaphors. You would not say "mongo aids" or "mongo alzheimers" or "mongo diabetes." Even if it is common, it is not particularly useful. Metastasis is better.
Reality!=Market consensus (startups that lost their data using the first versions of mongo)
More often: HN consensus=Reality
At least ES had the decency to tell you it might lose your data.
- Revenue: Total revenue was $150.8 million in the third quarter fiscal 2021, an increase of 38% year-over-year. Subscription revenue was $144.1 million, an increase of 39% year-over-year, and services revenue was $6.7 million, an increase of 19% year-over-year.
- Gross Profit: Gross profit was $104.7 million in the third quarter fiscal 2021, representing a 69% gross margin, compared to 71% in the year-ago period. Non-GAAP gross profit was $108.6 million, representing a 72% non-GAAP gross margin.
- Loss from Operations: Loss from operations was $58.1 million in the third quarter fiscal 2021, compared to $38.7 million in the year-ago period. Non-GAAP loss from operations was $16.0 million, compared to $14.3 million in the year-ago period.
- Net Loss: Net loss was $72.7 million, or $1.22 per share, based on 59.4 million weighted-average shares outstanding in the third quarter fiscal 2021. This compares to $42.4 million, or $0.75 per share, based on 56.4 million weighted-average shares outstanding, in the year-ago period. Non-GAAP net loss was $18.2 million or $0.31 per share. This compares to $14.6 million, or $0.26 per share, in the year-ago period.
- Cash Flow: As of October 31, 2020, MongoDB had $966.8 million in cash, cash equivalents, short-term investments and restricted cash. During the three months ended October 31, 2020, MongoDB used $8.1 million of cash from operations, $5.6 million in capital expenditures and $1.2 million in principal repayments of finance leases, leading to negative free cash flow of $14.9 million, compared to negative free cash flow of $13.1 million in the year-ago period.
For a dedicated database that has 32 GB of RAM, 8 virtualised CPUs, and only 160GB of storage you will pay $2 an hour! That's $17,520 a year for a database with the same level of compute and far less storage than my laptop.
I realise a big part of their target market is firms who are convinced they either can't hire skilled sysadmins or are convinced their time is worth more than the US President but at $17,520/yr for a tiny database it only takes two or three of those before you could, in fact, hire a full time sysadmin in a cheap location. And there's no way that maintaining these things would require an FTE, it's not even close to being a full time job. So how do the economics of this work out? Can anyone explain to me?
My best guess is that Atlas customers are companies that are totally price insensitive and have some sort of "everything must be cloud managed without exception" rule imposed from the top down. As Mongo is licensed such that only they can run a hosted service, such customers would find themselves forced by internal policy into paying whatever price Atlas charges. This would also explain their very low services revenue. But I don't know that, it's just a guess.
Idk, but maybe the engineers don't fully trust it, so they outsource the liability of running it with some contractual guarantees. And at the moment, this probably looks more fashionable, so no one dares question it in fear of seeming backwards.
Its not a contradiction to dislike mongodb technically well still thinking its a good business.
There was a large discussion (948 points, 267 comments) at the time: https://news.ycombinator.com/item?id=13421608
The first one, an object-oriented database company (last 80s, early 90s) suffered from an extreme case of misperceived markets, due to the occasionally intentional blurring of the lines between a "database system" (which means SQL to nearly everyone, and we didn't to SQL), and a persistent storage class added to C/C++ (and later Smalltalk and Java).
The second one (mid 2000s) was a neat physical storage trick, which should have been a feature of a database system, not an excuse for building a new one. My bad for missing this, but I really, really liked the physical storage ideas, so I joined.
Building and selling new database technology is extremely difficult. There are successful, huge, entrenched companies. There is, therefore, no good reason for anyone to gamble on your new technology. Your only chance is to convince the architecture astronauts at some prospect that they just have to have your product, but the odds of such a decision sticking all the way through the delivery of their product is extremely low, no matter how good your technology is.
Can you say what database this was? Akiban? SchoonerSQL?
I wish they didn’t run out of money and give up. In many ways startups feel like you gotta be a cockroach. The goal is to not die and live long enough to be profitable where you control your own destiny.
Like mongo atlas, rethinkdb’s hosted solution (horizon I think) could have easily taken off.
Like Firebase’s Firestore, it could have been quite successful.
MongoDB and ArangoDB did well so there seems to be a valid case for new databases. Rethink was loved by HN commenters. I used it in a startup I was working at in 2015-2017 and it was a pleasure.
developer tool market seems to be the only one I'm familiar with as an engineer. If I were to start a business, I will probably also do something for this market.
In general, it's much better to be in a market where you customers cannot do you your job, not just one where they don't want to do your job. In the latter markets, they start asking "Why should I pay you to do something I can do myself?", or just start doing it themselves, and that holds down the price you can charge.
This applies to a lot of familiar service businesses as well: housekeepers, gardeners, delivery/taxi/rideshare drivers, childcare, etc. Think of how much a doctor gets paid vs. a nurse vs. a home healthcare worker. The difference is all in skills that the provider can do but the customer cannot.
Fundamentally, business is about specialisation. Your customers always have the option of becoming jacks-of-all-trades but generally recognise it's best not to. Obviously if the value you're selling is so thin that your customers regularly and genuinely feel they could roll their own without incurring much cost, that's no good, but that's going to typically be the case in the early days before you had a chance to build up much value.
So it is with companies bringing core competencies in-house. Each step takes time, and is fraught with risk. It's not a matter of simply hiring people with the right skillsets. You need to know what the right skillsets are, and how to judge them. You need to hire managers and executives to oversee them. You need to convince all of the above people that they should work for this new company that's just entering the market and may reconsider in a couple years, vs. stick with their bread and butter that's been doing it for decades. Company leadership needs to prioritize integrating the new competency, which often means challenging a lot of their assumptions about how the business works. Investors need to be on board: integrating your suppliers often does bad things to your margins and costs. Tooling and capital investments need to be made, and they can take years to plan out and execute.
With developer tools specifically, the alternative to buying the product is often "We'll have a team of engineers spend a quarter or two building something that is specific to our needs", and no changes need to be made at the management/executive/personnel/financial levels. If, say, you're building a SaaS for a non-tech business, the alternative is "We need to learn how to manage software development processes, and hire a bunch of engineers, and get executive buy-in, and re-build the communication pathways in our org so this new division knows what to build." A lot of businesses actually try this, but the results have been predictably bad, and that creates a ready market for non-tech SaaS vendors.
In practice one company acquiring another, or even just poaching some employees, is a common thing that happens so frequently we take it for granted. When Apple acquired PA Semi nobody batted an eyelid even though they were in-sourcing something as complex and difficult as CPU design. And there are many examples outside of the computing industry too.
With developer tools specifically, the alternative to buying the product is often "We'll have a team of engineers spend a quarter or two building something that is specific to our needs"
If it takes a small team of engineers only six months to roll their own version of your product, that's what I meant by the value proposition being quite thin (or your price being too high). Also, most companies outside of very rich tech firms will not actually staff up an entire project just to clone a tool they could acquire off the shelf, unless that really is the only reasonable path forward e.g. their needs genuinely are unique.
Now, in computing we do have the issue that the prevalence of VC money, and indifference of VCs to whether it's being well spent, has created a large population of "very rich tech firms" with lots of apparently bored engineers working at them. So they spend a lot of time churning out open source frameworks that they maintain for a few years and then abandon, even when they could have used something else. That's rather unique to software but I don't think it's something inherent to developers as a type of person. Rather, it's inherent to a market where money is free and firms compare themselves to each other by the sheer size of their engineering teams. Outside of the Valley that problem mostly goes away.
Another analogy: Tesla's customers get the idea to just design and build their own cars instead.
Thanks for an insightful reply -- slightly helps me when thinking about my own projects
Yeah, but those are outliers drowning in a sea of free development tools. The price that they can charge is effectively limited by how much it would cost some to replicate the subset of that tool that they actually use.
A business that is paying $X/year for $DEVSOFTWARE for their entire devteam pays that only while $X is less than the cost of paying for developing the subset of $DEVSOFTWARE that they use. That limits what Atlassian can charge.
Compare to selling a tool to accountants - it doesn't matter how much $ACCSOFTWARE costs, the accounting department will only shop around, they will never just write their own (outliers excepted, of course).
I agree though that you should never sell to developers, most of them really don't care how they spend their time as long as they are occupied and work on something they deem interesting, so they will happily spend months or years reinventing existing solutions if given the opportunity (I've seen this again and again). You need to sell to the people that run the company and pay the developers or to the engineering management, because they know how much their developers cost, and they understand ROI better than they do. That said often developers are quite creative when making up reasons why not to use an external solution or tool and instead write it themselves, so whether you succeed in selling to a company depends on who's ultimately in charge of technology.
I worked for one and you're right, it was madness. They still did it though.
Things like a database would have to go through procurement if I wanted to use them and they weren’t free, and I avoid procurement like the plague.
Why the hell wants to spend their time justifying why something is the best solution to people that don’t have a clue.
I’d only go through that if the difference in quality was so palpable that I basically had no choice.
'Developer tools seemed like a good industry to be in, "sell shovels in the gold rush" and all, but it turns out developers prefer to dig for gold with their teeth.'
https://twitter.com/andygocke/status/1017509689695715328?lan...
But for the company I work for, I don't care. I would recommend my employer the fastest and easiest tool. And my employer usually don't want to spend engineering time on reinventing wheels.
If I were to launch a business in this market, the targeted customers can't be individual developers, more likely other businesses.
Like gihub, individual developers who only use their free account are part of github's marketing team.
I'd probably word it differently:
Developer tools seemed like a good industry to be in, "sell shovels in the gold rush" and all, but it turns out developers prefer to make their own shovels to dig for gold (because it is often cheaper than buying).
It’s a colorscheme, yes, but you pay for the consistent support of over 160 apps, which is a big deal IMO. Heck, every program under the sun seems to offer a solarized dark theme, but they’re all wrong in different ways and use different accent colors for different things, so I rolled my own in a lot of cases to get an actually consistent appearance. It’s a huge hassle, and just paying 80$ would probably have saved money compared to the hours and days I spent manually theming stuff.
Take the internet for example - before the internet existed, the 'market' for the internet sucked.
For me, any company that failed due to solving too ambitious of a problem is the best kind group of people to invest into again and again until they succeed and succeed big. The people running it will have learned the easy lesson of solving a smaller problem faster, while retaining the chops to actually solve hard problems.
All too often, people fail because they solve too small of a problem - there was a recent discussion about that on here regarding a macos calendar widget.
Open source means that all value and good ideas turn into commodities. Whatever mongo did ten years ago that set them apart is now a commodity. You can get it from dozens of different OSS products that have since built similar features or that have been created since then. That does not devalue mongo's product and they seem to be making some nice profits monetizing it. It's just that their value proposition has now shifted to hosting and support.
RethinkDB as a database actually did not fail. It's just the company behind it that failed. And strictly speaking, it merely failed to deliver on unrealistic expectations its investors had. Companies fail all the time; usually it has to do with product market fit, figuring out a sane business model, etc. However, the database still exists. People still use it. People still work on it. I've never used it myself. But it looks like an interesting product.
Whining that people won't pay for a database is pointless. There are plenty of successful companies making money from their OSS software. Most companies don't get to be unicorns though. That's disappointing if you are a VC but otherwise not even a goal for many healthy companies that are just looking to make healthy profits doing honest work.
In this case, VCs helped fund an unsuccessful company that happened to have created a successful product. That happens a lot in that world. VCs pay for people to take risks and sometimes that just doesn't work out and they walk away from it. But because it was open source, they are the only ones that lost. Mysql has seen many companies come and go but the product seems to stick around. Same for Postgresql. Both have a very healthy ecosystem of large and small players depending on and making money with these products. Some of these are unicorns even.
To break out of "developers making tools for developers" you need to get some domain knowledge.
Stuff I don't pay is stuff I feel I can do better if I had some free time and rather gets in the way than properly solve problems and stuff that tries to charge more than what they're worth which can be said for many online developer services.
Don't expect to make big bucks selling to single devs.
However... who you sell to isn't the people giving you GitHub stars and retweets, and they will have diff needs. if you lose sight of that....
I don’t think something like VSCode should be free, or a single api call on any platform, etc. Absolutely nothing should be free because people get too used to the idea of free. It’s a human failure.
You have it confused - prices is an arguably necessary, temporary condition, until we get back to simply from everyone according to their ability, to each according to their needs.
Even if there was a society like that, I don't think I'd want to go - not only did they go extinct, they don't have mRNA vaccines, spacecraft, and Telemundo soaps.
In theory, sure, sounds great, but the pragmatic timescale it would take to get there would be beyond lifetimes. We’ll all be dead by the time these systems reach their true form.
In your particular example for instance, human nature would wreck havoc in the form of abuse. I will point to (incoming snark) the very rare cases of slavery, underpaid labor, serfdom, like you know, these rare things that happened through the centuries. So yeah, I suppose that idea is ‘still working itself out’ two thousand years later.
Fuck all that is my point, as many souls have been lost waiting for utopia. Ask the guys that built the pyramids.
What do children pay their parents, or parents pay their kids? That's the relationship people start out with and one that is most natural.
People seem to think that there is money, and before money, there was barter. That's fake news made up by economists (the barter before money bit).
People have been around for hundreds of thousands of years. Money has been around for a few thousand in select places and never used for most transactions by most people until extremely recently.
I'd argue it's still not used for the most important transactions and the existence of food stamps, universal health care in every sane developed country, free education and subsidies for parents with children is simply some of the numerous examples of most people preferring free for things they deeply value and care about.
Money is great for veblen goods and exchange among adversaries, everything else wants to be free. Free as in decided based on factors other than money, not free as in oxygen.
I'm not articulating this too well, please have a look here for a take I'm trying to express, done by a proper journalist.
https://www.theatlantic.com/business/archive/2016/02/barter-...
If at any time people decide they no longer believe in home ownership, then the house I built for you had no value since no one will build my home when it’s my turn.
This wouldn’t be a problem if you just paid me cash since it stores value.
We can come up with pro and con examples for any set of ideas.
I remain of the mindset that monetary transactions within a community are a net negative, largely because I think human dignity should not be something you can purchase, sell or bargain for.
If we provide human dignity for every member and let money be something we use for nice-to-have items/services, that strikes me as a significant improvement.
It's been known as utopian since at least the time of Thomas More[1].
> What do children pay their parents, or parents pay their kids? That's the relationship people start out with and one that is most natural.
Yes, it's natural for that relationship. That does not mean that:
a) it applies to other kinds of relationship
b) it scales to all of society
In fact, it's that kind of thinking that leads to the authoritarianism of Confucianism, the tyranny of communism, and the stifling paternalism of western social democracy (among other things such as outright fascism).
Gifting worked amongst tribes because they're all related. As to Graeber, he's saying that slavery is the creation of money? No, he's too mealy mouthed for that “perhaps not creating but at least enabling institutions such as slavery”.
Wow, money enables slavery. What a pathetic attempt at insight and guilt by association.
Money is simply a better system than the competitors, as is liberalism - the wish that people be free to choose how they use their spending power (or gifting power, ha) - to its competitors. A gift economy was not able to lift almost the entire world out of poverty (and it will hopefully complete that effort soon[2]). Communism and socialism (not that there's a real difference) have only managed to emiserate and imprison every society it's been tried in. No thanks.
how is it bad when snowflake just had the biggest tech ipo of all time.
That’s not a reason for failure, of course, but I have mentioned before that we (as in developers) often only choose things with permissive licenses, often to the detriment of the products we “support” (see the recent elastic drama, where we blamed elastic being forked by Amazon for being too permissive!).
What would you say is the problem with AGPL? I was thinking of releasing a product under it myself.
Huh?
(Related: I've worked with new managers in SV who couldn't define what the words "leadership", "responsibility", "planning" or "accountability" meant, since all they knew was PHP programming.)
It's a good post mortem read.
I gave the same advice about "use case" as in the post mortem to a YC database company a few years ago. The "founders" also thought they were building a better (technical) mousetrap, but I told them nobody was going to pay anything if it did the same as an existing database - it had to do something unique and be cloud-based, like Snowflake, or RDS. They were crestfallen, but still didn't listen to me. I even got hate mail from them before they went under - talk about being clueless to the bitter end.
What RethinkDB did wrong was what was in their post mortem, but also not having experienced DBAs to review the use cases. As a DBA, my default is to say no to new databases, and being the audience, that would have been valuable for them to know.
An example of how DBAs and devs think differently is the Jepsen database reports - devs fetishize those, but DBAs are like, "We already know corner cases aren't going to work most of the time. What's your point exactly?"
Source: experienced DBA.
What more reason is needed?
Not at all. All of MongoDB is just a feature in Postgres, and Postgres does it better, and Postgres does hundreds more things too.
The idea that PostgreSQL is a superset of MongoDB is incorrect. As some examples, PostgreSQL doesn't shard, have multi-region, multi-cloud, enforce JSON schema, data governance, full JSON support, etc.
That said, PostgreSQL does do a lot of things MongoDB doesn't do.
The two products have a big feature overlap - both are great databases. I think you could build a modern app on either one. If you want to build in modern languages faster and have a more available system out of the box, I'd suggest MongoDB. If you have legacy skillsets, legacy apps, or just love relational, I'd suggest PostgreSQL.
I don’t disagree with you, but the people that choose Mongo do not see postgres as a valid alternative.
being hip, which is one of the most important factors for choosing a database database technology, is not one of them though.
Both founded by Keith Bostic, from Berkeley DB heritage.
This article has less to do with databases or open source software than it does with the fundamentally misaligned incentives of egalitarianism and capitalism.
I wish there was a social open source license that said "you can use this indefinitely as long as you pay us something". The price would be up to the user but the generally accepted polite minimum would be at least a penny ($0.01).
Then businesses could publicly display their level of support (both initially and yearly) for the software they use, to attract customers. They could also be audited by the IRS, so a lack of patronage could correlate to a lack of equality and maybe even reveal corruption and other malfeasance. At the very least it would reveal a lack of internal controls.
This could work kind of like UBI for open source. And I'm definitely not the only one who has ever thought about this. But there is just so much free capital floating around right now that it's a great time to be thinking about how to reform the paradigms that we all depend on.
Maybe another, important, lesson the RethinkDB team left out was: raise more capital than you need, and go public fast.
Two years later, MongoDB the company is now valued at $19.491B. Like many other public tech companies, it is yet to turn a single profitable quarter, and is not expected to in the next few years.
Annual YoY revenue growth is flat too at around $500m. Equity is negative.
Perhaps because the market thinks useful product companies will be valuable vs. the US dollar in the future. And its a vital part of many companies infrastructure these days, without much of a solid enterprise alternative.
If anyone tried to sell git itself it would be impossible. I never used mongodb for reasons listed in the article. But there is free and open source and resilient postgres.
the companies who succeeded around git did not build git. it would be a tough thing to develop both git and the hosting / workflow business around it.
so I guess the lesson to learn is don't try to build extremely sophisticated software as a startup where there are already good enough open source alternatives.
The industry is still in an awkward position where software deals in 'intellectual property', but the concept was never really developed with an understanding of whatever it is that software is. Something like PostgreSQL for example isn't exactly a product, a service or a novel idea. Most of its consumers don't use most of its features. It is almost a perspective on a problem and some codified good design ideas.
It isn't obvious if selling a perspective is profitable. There are a lot of winners in software that don't actually sell software. Eg, Facebook/Google sells eyeballs, Amazon sells infrastructure, Apple sells iPhones.
MongoDB.Inc probably doesn't sell 'a database' if the details of their customer relationships were open for inspection.
There is an archive of his deleted tweets here where he describes it: https://gist.github.com/travisbrown/059310042193a2e143408b05...
And then there was an article that seems to be gone from the web (https://web.archive.org/web/20201130215752/https://threader....). It stands to reason this huge distraction played a role in the mismanagement of the database and why it ultimately didn’t succeed.
That said, I'm wasting time on HN currently. But I'm also on holiday.
Personally, I appreciate the frankness. Too bad it got deleted.
Uh no, It was slow on real world workloads, you just don't want to admit it.
Read The Economist religiously. It will make you better faster.
The challenge in fast changing markets is that the best information isn't written down anywhere (and especially not in magazines!), but that it's locked inside people's heads.
At least, this is what has worked for me (TimescaleDB founder)!
Also, generally the Financial Times is the better standard for strictly business related journalism AFAIK, maybe they're neck-in-neck.
Spending 30 minutes to learn from someone in your field about how to get your next lead or which part of the internet you should be advertising to is a much better use of your time.
Just curious on what’s the best way to do market research for the product
Just curious on what’s the best way to do market research for the product
But also good to "sanity check" your instincts by checking in with people whom you respect and whom you think will understand the market better than you (which is what my parent comment is trying to get at).
I would add to your list: spend as much time as possible with customers and figure out the actual pain.
(In fact, one of our company values at TimescaleDB is, "Be a student in both work and life".)
BTW - who is this? My apologies, I don't recognize the username :-)