Absurd Success
marginalia.nu
marginalia.nu
It makes me deeply happy to hear success stories like this for a project that's moving in the correctly opposite direction to that of the rest of the world.
Engildification. Of which there should be more!
My soul was also satisfied by the Sleeping At Night post which, along with the recent "Lie Still in Bed" article, makes for very simple options to attempt to fix sleep (discipline) issues.
The author recently went full time on their Marginalia search engine, AFAIK it's a team size of 1, so it's the farthest away from any enshittification risk. Au contraire, like you say: it's at these sizes where you make jewels. Where creativity, ingenuity and vision shines.
This comment is sponsored by the "quit your desk job and go work for yourself" gang.
A bootstraped company can resist enshitification indefinitely, but if it gets any investment, resisting it becomes harder and harder up to the point where a publicly traded one can't resist it at all.
It could be it’s resolved though funding them, or getting more people across it. Anything to help mitigate the risk.
Or don't, I guess the world needs its mindless drones, that discourage everybody else from taking the leap and following their dreams.
I joke, but a lot of people just want to show up and do the bare minimal until quitting time. Disturbingly, many seem employed in customer facing positions.
To be fair, if they’re paying minimum wage, the employer has the same attitude towards the employees.
- cut his resources burning in half,
- is more productive with a smaller screen than before, and
- sleeps like a log at night
(his 3 last blog posts!)
Too much processing power, too much memory and storage. Too many data centers, bandwidth. We lost all sense of traction and we're running wild trying to fill up all the extra space and burn up all the extra resources. Software is currently way more complicated than it needs to be, and it makes it overall excessively insecure as well.
I've said this before, and it's becoming official this fall. I'm getting out of the wheel. I wish I had the competence of Marginalia's author to brave this storm. But my vessel is making water fast and I need to make landfall. Next month I embark on a new stint in academia hoping to change direction for good.
There might be an argument that education has never been the primary responsibility of the academy; that it's institutional responsibility is on "learning" not "education." If your goals are primarily research oriented there's still room to do that, especially in STEM.
Or tiling window managers with virtual workspaces. When most applications I use are two keystrokes away, the need for more than one screen seriously lessens. Besides, looking at a screen on the side hurts my neck.
I've had 3 monitors at one point on my main dev box of which 2 were 32" 4K. It became tiring very fast. I went to a single one. First it was 40" 4K but it proved to be too much neck bending to look at. So I've finally settled to a single 32" 4K and am happy (I use it at 100% scaling). Smaller monitors - no thanks, not my cup of tea. I know it is possible and I used to program on tiny ones in my young days but fuck it. Do not want to go back.
In terms of size I use 28" and that works well for me, 14" on the laptop is fine too but feels a bit limited now and then. In the long term I'll probably go for something inbetween with a size ratio closer to 4:3.
I found something with an effective resolution of about 2560x1440 to be a sweet spot, where I can fit windows needed for a single task in one screen. Provided those windows don’t waste too much space with unnecessary stuff of course.
I quickly realized I absolutely hate setups with multiple monitors, when I was given one at work. I quickly gave back my second monitor, because it just made it worse for me, and additionally wasted space on my desk. If you want to improve my circumstances at work, give me a bigger monitor, not more of them.
(I've never even tried multiple monitors. I could immediately think of several reasons why they would not be an improvement for me, so made a quick decision in 1998 and moved on.)
I’ve been struggling with sleep this year and finding out what works for others is very useful. I wouldn’t have found it if not for your comment.
Link for others interested: https://www.marginalia.nu/log/86-sleep/
Having said that, I did look up 'gild' to make sure I was using a semi-obscure word at the opposite end of the scale that begins with 'shit'.
Coming up with new words is surprisingly easy when there's an existing word requiring of a foil.
Don't give me too much credit for this, I need to maintain my current level of almost total obscurity in order keep pumping out the quality ;)
> Engildification. Of which there should be more!
Great word though.
https://help.kagi.com/kagi/search-details/search-sources.htm...
If you use the "non-commercial" lens, those results, among with results from Kagi's own index and a few other independent sources will be prioritized.
I'm wondering if humans are mostly incapable of producing great things without (artifical) restrictions.
In this case, marginalia is (ridiculously) efficient because Victor (the creator) is intentionally restricting what hardware it runs on and how much ram it has.
If he just caved in and added another 32GiB it would work for a while, but the inefficient design would persist and the problem would just show it's head later and then there would be more complexity around that design and it might not be as easy to fix then.
If the original thesis is correct, then I think it explains why most software is so bad (bloated, slow, buggy) nowadays. It's because very few individual pieces of software nowadays are hitting any limits (in isolation). So each individual piece is terribly inefficient but with the latest M2 Pro and GiB connection you can just keep ahead of the curve where it becomes a problem.
Anyways, turned into a rant; but the conclusion might be to limit yourself, and you (and e everyone else) will be better off long term.
For most applications it simply does not make any sense to spend this much time on relatively small optimizations. If you can choose to either buy 32GiB of RAM for your server for less than $50 or spend probably over 40 hours of developer time at at least $20 / hour, it is quite obvious which one makes more sense from a business perspective. Not to mention that the website was offline for an entire week - that alone would've killed most businesses!
A lot of tech people really like doing such deep dives and would happily spend years micro-optimizing even the most trivial code, but endless "yak shaving" isn't going to pay any bills. When the code runs on a trivial number of machines, it probably just isn't worth it. Not to mention that such optimizations often end up in code which is more difficult to maintain.
In my opinion, a lot of "software bloat" we see these days for apps running on user machines comes from a mismatch between the developer machine and the user machine. The developer is often equipped with a high-end workstation as they simply need those resources to do their job, but they end up using the same machine to do basic testing. On the other hand, the user is running it on a five-year-old machine which was at best mid-range when they bought it.
You can't really sell "we can save 150MB of memory" to your manager, but you can sell "saving 150MB of memory will make our app's performance go from terrible to borderline for 10% of users".
It might just be to a certain degree, we’re not actually getting any business efficiency from creating bloated and slow software?
A lot of things, especially in business IT, are built on top of outdated and misleading assumptions and are leaning on patterns and norms touted as best practices.
We sometimes get trapped in this belief that any form of performance improvement somehow costs us something. What if it’s baggage that we didn’t need in the first place?
The entire project a fractal of this type of business-inscrutable engineering, and in any organization where engineers aren't calling the shots, that engineering isn't going to get done, and the project is going to be hideously slow and expensive as a result.
In a parallel universe where I had gone the classic startup route and started out using the biggest off-the-shelf pieces I could find, gluing them together with python (instead of building a bespoke index); then thrown VC money at hardware at when it started struggling; even more when it struggled again; I'd be absolutely dumbfounded when it yet again hit a brick wall and my hardware cost is tens of thousands every month (as opposed to $100/mo now).
Since I've instead built the solution from scratch, I've also built a deep understanding of both the domain and the solution, and when I'm faced with scaling problems, they're solvable in software rather than hardware. I can just change how the software works until it does. It's a slower route, but it's also able to take you places where conventional-wisdom-driven-development does not.
But it does always cost you something. Developer time isn't free, after all.
If we only cared about performance, we would be handwriting SIMD intrinsics for baremetal applications. But we don't, because it is easily worth a 20% performance penalty to write code in a modern programming language. We're willing to trade 10% performance for a well-tested framework and library ecosystem which greatly reduces development time. Nobody cares how efficient your application is when it never ships.
Even "bloated" and "slow" do not always mean the same thing. Just look at something like database indexes: they take up space (bloated), but make your application faster. Often that's a worthwhile tradeoff, but creating indexes for literally everything isn't a good idea either. It is all about finding the right balance.
I do agree that a lot of user-facing applications have gone way too far, though. Even completely trivial Android apps are 150MB+ binaries these days, and the widespread use of memory-hogging Electron tools is a bit worrying. When your app runs on millions of devices, you should care about resource usage!
I think this is the sticking point. People assert without any real evidence that whatever framework greatly reduces development time, and if that were the tradeoff, it might make sense. e.g. Rails and Laravel bill themselves this way.
Meanwhile, I've found that a more barebones framework in Scala is more productive to develop with, and also gets at least 100x the performance (e.g. a request rate in the 10s of thousands/second is easy to do on laptop hardware), which also makes it operationally easier since now you don't need a giant cluster.
According to the rest of the comment we seem to largely agree. But this statement is what I want to challenge a bit.
Basically it comes down to this: We're doing plenty of unnecessary stuff all the time, especially in business IT (web applications, CRUD, CRM, CMS, reporting etc.) software.
Your Electron example points to the right direction, but there are also a lot of practices regarding code organization (patterns, paradigms), being too liberal with pulling down dependencies, framework bloat etc.
Simply getting rid of things that we don't need and thinking about performance in very rough terms gets us _both_ better performance and minimal code. I would wager that this isn't a tradeoff as in you describe. We might, especially in the mid-long term, actually gain developer time.
This often means thinking about DB access patterns, indexing and so on as you mentioned. Meaning we think in SQL and lean on the capabilities and heuristics of our DB. What does the DB actually need to do when I do these queries? How can I model the data in a way that gives me just enough flexibility with reasonable performance? Which parts of the system needs to know what? How does the data need to flow through it?
All that stuff we sometimes put on top (ORMs, OO patterns etc.) can get in the way of that. Does it really make us more productive to put these abstractions on top? Do we gain anything from doing these OO incantations?
The article in question is a really good example of removing complexity and drastically increasing performance at the same time.
I have a good example as well.
Our internal image optimization module that we use to generate differently sized images and formats in order to accommodate browser capabilities and screen sizes, was getting noticeably slow, especially for e-commerce related sites and brochures that typically feature a gallery per product/service.
Long story short: It got 50-60x faster simply by removing a convenience layer, writing SQL directly, processing the images in grouped batches and so on. AKA all just low hanging fruit. The end result is also simpler. It took work/time but it didn't need to be that slow in the first place.
And we are our own users too! We have to maintain, extend, test our code. Faster code gives us faster iteration cycles. Fast feedback loops lead to much higher productivity and flow.
Things normally don't exist in an optimum state where you must necessarily lose one thing to gain another.
Incidentally, I think the reason they need those specs is the same: the people building the dev tools all have top end hardware, and what’s fast enough for them is good enough to ship. I don’t think the people building the dev tools at meta, or apple, or google are seriously considering the use case of a developer working on an old dual core 8 gb machine, but that’s the reality in large parts of the world.
On the other hand, development tools simply require a lot of resources: a debug build is always going to be more demanding than a production build, and a full recompilation of something like Firefox is never going to be fast.
I reckon a lot of developers are more than happy to sacrifice 8GB of RAM for near-instant advanced autocomplete and typechecking, but we should keep in mind that it should not become a requirement to do basic development.
I will keep using the heavyweight development tools that run through all of my code discovering problems, analyzing its quality, changing it into more performant code, deducting all kinds of new code to complete my high-level view, and doing all kinds of the modern benefits that use my computer to replace my work.
I can get behind testing the resulting executable on a low-power machine. But not on using it for everything.
SO here artificial constraint is that "you cant have more RAM" and so you need to find other more creative solutions.
Working on improvements like this makes a lot of sense in academia, when you are running thousands of servers, or when you need so much memory that you can't buy a big enough server. It's a nice backlog item for when there is literally nothing else to do.
I have definitely fallen into this rabbit hole myself. Solving difficult problems with clever solutions is a lot of fun! But my manager - rightfully - chastised me for it because from a business PoV it was essentially a waste of time and money. Fine for a hobby project, not so much when you are trying to run a business.
It also makes sense for all desktop and mobile apps, per gnyman’s point on not hitting individual limits. Also for anything that other people will be running, like libraries. Because your thing might perform marvelously in isolation on a beefy dev machine and still run like ass on every user’s somewhat underpowered computer with ten other apps competing for RAM (hello Slack).
See also: “craptop duty”[1]; how the 512K Mac reduced the competitive advantage of MS’s ruthlessly-optimized office apps but the MultiFinder, né Switcher, restored it[2].
[1] https://css-tricks.com/test-your-product-on-a-crappy-laptop/
[2] https://www.folklore.org/StoryView.py?project=Macintosh&stor...
And yes, hearty recommendation on the craptop.
I was discussing with a colleague a while back when he said he had to get one as he was unable to reproduce a problem on his M2 even with the 6x cpu throttling you can do in Chrome. On the old laptop he dug out of storage he managed to reproduce it instantly.
I need to check with him how easily he could fix it, one argument I have to why some optimisation is always good is that it's easier to fix things as you go along vs having to go back and try to make things better. I mean if there is one thing making it slow then it's easy, but I have a feeling it's often a mix of things and at that point you might not be able to replace the slow parts as everything else depends on them.
I think maybe you dropped a zero.
In reality the listed changes once you consider all team members are likely going to run closer to 80-120 hours at $150 / hour - but that only confirms the final conclusion.
It's actually how we used to build software. It's why we could have an entire operating system perform well on a machine like a Pentium 1 with most of what you'd expect today, etc. while at the same time we have web pages that struggle to scroll smoothly on a smartphone with literally a thousand times more resources across all axes. The Word 95 team were constantly faced with limits and performance tradeoffs, and it very clearly worked or did not.
If I had just gone and added more RAM (or whatever), I would still have been stuck with an inferior design, and soon enough I would need to buy even more RAM. The crazy part about this change is that it isn't just reducing the resource utilization, it's actually making the system more capable, and faster because free RAM means more disk caching.
[1] e.g. this runs on a single pi, and is much faster than production wikipedia because it doesn't permit updates: https://encyclopedia.marginalia.nu/article/Hacker_News
I think it’s why old computers felt good and also why old games were so good.
Maybe it have something to do with the complexity of the systems we deal with.
When you have a restricted amount of some resource (RAM, physical space, food, materials, time, money …) you have to plan how you will use it. You are forced to be smart.
When you have a virtually infinite resource, you can make whatever you feel making but you don’t have to really care about the final state, you just start and you’ll see when it will work.
I’m not exactly a true gamer, but I’ve always been amazed by the fact humans were capable to store so much emotion, adventures and time to enjoy in the good old cartridges with some kb/mb of rom. I mean, Ocarina Of Time rom is just the size of the last 8 photos I took with my iPhone.
A raspberry pi 2 can do over 4 billion Dhrystone instructions per second, and a pi 4 over 10 billion per second.
Of course by modern standards mid-1970s SABRE was pretty barebones for an airline's main system, but it's at least theoretically possible to run simplified systems for over a 100 airlines simultaneously on a single pi 2...
So yes modern programs are very far from optimized. 1000x or 10 000x improvements are possible, less for math heavy stuff.
I think microsoft has a huge problem with this. Even 3000$ laptops from 5 years ago struggle with running a teams call, some office instances and a browser with 30 tabs at the same time without slowing down to unacceptable levels.
They test stuff individually and running one thing alone is fine, but that's not what people do.
I'd imagine that artificial limits in the form of run time on well-defined hardware that are only raised after an explicit decision could be the solution to this.
But then again I only write business software where the performance aspect comes down to "don't do stupid shit with the database and don't worry about the rest because the client won't pay for those worries", so I might be on the wrong track entirely.
When I see a website...
- using a ridiculously thin or small font,
- relying on a high-end monitor to provide sufficient color contrast
- loading an unreasonable amount of resources only to provide laggy animations
...I'm wondering if the responsible designer(s) only have 32-in Retina displays and the latest Macbooks to work with. Because on any other combination of devices, the website looks and feels awful.
And I know this because I was formerly guilty of it!
I think often it’s that they aren’t users themselves. They make it “pretty” but not functional.
I was able to spot obvious flaws in the design, she agreed but said that she had no time and such is life.
On the one hand, I agree with this. I think an awful lot of great art and great work comes from the enforced genius of operating within constraints, and there's a profound feeling that comes from recognizing that kind of brilliance.
I'll also say, though, that there's also something about seeing the results of absolutely turning every nob to 11 - about seeing the absolute unfettered apex of what particularly talented people can actually do with no constraints whatsoever. It's a very different experience, and I deeply respect the genius of making art under constraint, but sometimes you've just gotta put a dude on the moon, you know?
in software development it looks like everyone and their grandmother is sending people to all moons known to mankind, though.
And crash landing every time. To do a far reaching soft landing you need to have learned to reach LEO within constraints.
Here is an opinion to support your hypothesis from a couple of different domains: poetry and music.
While some people prefer free verse and avante-garde music, what stays most in my mind, and what seems to endure longest overall, are poetry with regular rhyme and meter and music that follows standard patterns of melody, rhythm, and harmony. Having to force their creativity into those sometimes rigid frameworks seems to enable many artists to produce better works.
I write software in ABAP, which is a weird and ridiculously complicated language that has inline SQL and type checking against the database and has never had a major version that breaks old stuff, so code from 30 years ago will (and does) still run.
I used to have fun working around the quirks of it and finding solutions that work within the limitations, but now I'm just frustrated by having to solve problems that haven't been around for the past 15 years in the rest of world and looking at terrible code that can't be made any nicer because of those limitations or because customers don't give a shit as long as it works most of the time.
And as a developer or a team, you're bound by how long development takes, not by the required resources.
You won't be asked by a business stakeholder "oh, and how much RAM does it take?" or "why is it $2,000 a month instead of $1,000?". These questions tend to come much later when profit needs to be ironed out.
Further, it feels like there's a corollary here to companies, where financially constrained companies who are smaller and more focused provide better customer experience than cash-flush competitors.
I think the real issue is that there isn't a program language that produces a compiler error if the given code can exceed a maximum specified latency.
Even working on a program with soft-realtime scheduling, I've had to constantly push back against patches that introduce some obscure convenience without having measured worst case latency.
The problem is so bad I doubt most people realize it's there. I don't know what the answer is, but I have the feeling there's an intersection with timing attacks on software/hardware. Some kind of tooling that makes both worst case times and variance as visible as the computed CSS in devTools would probably help. Added to some kind of static analysis, perhaps devs to hack their way to decently responsive interfaces and services.
It's about specialization around the usage of few elements to achieve a goal vs a paradox of choice or going through common and known patterns.
Jams can be great for this and people realize they can work so much more efficiently and focus on the core of their idea.
Does Marginalia support "time filters" for search like past day, past week etc? According the special keywords the only search params accepted is based on years.
year>2005 (beta) The document was ostensibly published in or after 2005
year=2005 (beta) The document was ostensibly published in 2005
year<2005 (beta) The document was ostensibly published in or before 2005Is that actually true for SSDs? For raw flash it’s not, provided you are overwriting “empty” all-ones values or otherwise only changing 1s to 0s. Writing is orders of magnitude slower than reading, but still a couple orders of magnitude faster than erasing (resetting back to “empty”), and only erases count against your wear budget. It sounds like an own goal for an SSD controller to not take advantage of that, although if the actual guts of it are log-structured then I could imagine it not being able to.
> Is that actually true for SSDs? For raw flash it’s not, provided you are overwriting “empty” all-ones values or otherwise only changing 1s to 0s.
Maybe it depends. I wrote the driver for more than one popular flash chips (don't remember which ones now, but that employer had a policy of never using components that were not mainstream and available from multiple suppliers) and all the chips I dealt with did read and write exclusively via fixed-size pages.
Since SSDs are collection of chips, I'd expect each chip on the SSD to only support fixed-size paged IO.
Normally the controller will use a whole bunch of tricks (e.g. overprovisioning, buffering and reordering of writes) to avoid this type of worst case pattern, but that only goes so far.
1. Writable Unit: The smallest unit you can write to in an SSD is a page.
2. Erasable Unit: The smallest unit you can erase in an SSD is a block, which consists of multiple pages.
So if a write operation impacts only 1 byte within a page, the SSD cannot erase just that byte. However, it does not need to erase the entire block either.
The SSD can perform a "read-modify-write" type of operation:
- Read the full page containing the byte that needs to change into the SSD's cache buffer.
- Modify just the byte that needs updating in the page cache.
- Erase a new empty block.
- Write the modified page from cache to the new block.
- Update the FTL mapping tables to point to the updated page in the new block.
So, a page does need to be rewritten even if just 1 byte changes. Whole-block erasure is avoided until many pages within it need to be modified.
Not precisely. The logical view of a page living at some address of flash is not the reality. Pages get moved around the physical device as writes happen. The drive itself maintains a map of what addresses are used for what purpose, their health and so on. It’s a sparse storage scheme.
There’s even maintenance ops and garbage collection that happens occasionally or on command (like a TRIM).
In reality a “write” to a non-full drive is: 1. Figure out which page the data goes to. 2. Figure out if there’s data there or not. Read / modify / write if needed. 3. Figure out where to write the data. 4. Write the data. It might not go back where it started. In fact it probably won’t because of wear leveling.
You’re right that the controller does a far more complex set of steps for performance. That’s why an empty / new drive performs better for a while (page cache aside) then literally slows down compared to a “full” drive that’s old, with no spare pages.
Source: I was chief engineer for a cache-coherent memory mapped flash accelerator. We let a user map the drive very very efficiently in user space Linux, but eventually caved to the “easier” programming model of just being another hard drive after a while.
It's completely false. Even the most primitive SSD controllers would make some attempt at mitigating this.
Limitations of finance and hardware, making decisions about 32 vs 64 bit ids, sharding, speed of updating all sound very familiar.
Reminds me of Google way back when and their 'Google dance' that updated results once a month, nowadays it's a daily flux. It's all an evolution, and great to see Marginalia offering another view point into the web beyond big tech.
Lots of people treat optimization as some deep-black-magic thing[1], but most of the time, it's actually easier than fixing a typical bug; all you have to do is treat excessive resource usage identical to how you would treat a bug.
I'm going to make an assertion: most bugs that you can easily reproduce don't require wizardry to fix. If you can poke at a bug, then you can usually categorize it. Even the rare bugs that reveal a design flaw tend to do so readily once you can reproduce it.
Software that nobody has taken a critical eye to performance on is like software with 100s of easily reproducible bugs that nobody has ever debugged. You can chip away at them for quite a while until you run into anything that is hard.
1: I think this attitude is a bit of a hold-out from when people would do things like set their branch targets so that the drum head would reach the target at the same time the CPU wanted the instruction, and when resources were so constrained that everything was hand-written assembly with global memory-locations having different semantics depending on the stage the program was in. In that case, really smart people had already taken a critical eye to performance, so you need to find things they haven't found yet. This is rarely true of modern code.
Optimization is often an architectural problem. Sure there are cases where you're copying a thing where you could recycle a buffer, but you run out of those fairly quickly, and a profiler will tell you what you need to know.
A lot of the big performance wins are in changing the entire data logistics, possibly eliminating significant portions of the flow until the code does what it needs to in as few steps as possible.
SQLite has the benefit that it's a single file though, and you can do cool things with that. Such as copy it, share it, etc.
Of course I do respect the encapsulation of the existing SQL query basis but there is sometimes a time for something even more compact.
"I wish I knew what happened, or how to replicate it. It’s involved more upfront design than I normally do, by necessity. I like feeling my way forward in general, but there are problems where the approach just doesn’t work"
Yes, immediate (or soon enough) gratification feels good... To me, and maybe is because I am an old fart, this is the difference between programming and engineering.
https://search.marginalia.nu/search?query=encoding%2Fjson
but NOT what I am looking for. If I try again with Google:
https://google.com/search?q=encoding%2Fjson
first result is exactly what I want.
Though I strongly disagree with Putin's illegal war of aggression, I don't censor Russian websites. You'll find Russian truth on there as well: https://search.marginalia.nu/site/english.pravda.ru
Sure, we can shell out to libraries and to other people's work, but at some point, you will have to understand the thing that you've abstracted away if you want to evolve it.
Or as the kids say, let him cook: https://knowyourmeme.com/memes/let-him-cook-let-that-boy-coo...
It can be more or less translated to "I want to see the outcome of what they're doing" usually used after the person who is "cooking" is criticized for their methods.
I'm building stuff myself because it's the only way I'm aware of to run a search engine capable of indexing quarter of a billion documents on a PC.
Because you're able to offer a product with differences in the market.
Because you can.
It's not like they're implementing their own DB to get to a MVP of their product "Tinder, but like for dogs".
Well, funny you should say that, https://explore.marginalia.nu/init
But I'm a privacy and self-hosting nut, which is probably just another way of saying the same thing.
(I will probably never actually do it, but that doesn't stop it being on the list).
plenty of folks dabble in art, many of which quite poorly, when good and even great artwork can be purchased for not all that much.
some paint the walls of their house rather than call a professional painter.
there are countless reasons, including but not limited to the easiest, most flippant, and possibly the most human response to your question.
why not?
Because by not being a huge team you can do it better.
Read any articles about how amazing Google's search is lately? Me neither.
anybody or any team that sets out to build something will prioritize different things and end up with radically divergent implementations.