Hacker News Reader

Comment by stesch | Hacker News Reader

People are still using MongoDB for new projects after all the negative articles and comments about it?

Inertia aka traction is very powerful in tech. Consider that all these negative articles do not percolate to various semi-isolated communities and groups.

Unfortunately yes, but I'm not surprised.

There seem to be a large number of people who came into the tech-startup industry through programming a bit of Javascript, maybe from a web design background, who never had a formal education in computer science, who suddenly found out that they could write 'full' web apps in Javascript, and they could query a database in Javascript, but never learnt enough to know why this is a bad idea.

I'm not saying this is all bad, or that all Node devs are bad at all. Node is a good tool for a certain set of problems, and there are some excellent tools in the ecosystem and some great devs. But I think it is the easiest ecosystem to be a terrible developer in, and still get stuff working.

What does Mongo have to do with Javascript, and why is querying a DB with javascript a bad idea? What language you query a DB in should have no impact.

MongoDB isn't really "queried with Javascript". Rather, it's "queried with JSON", which is a separate language entirely.

MongoDB is bad because it has unsafe defaults (e.g., acknowledging writes that haven't happened), not because you can query it in JavaScript.

While I agree with most of your assessment, this is one bit I very much disagree with:

> But I think it is the easiest ecosystem to be a terrible developer in, and still get stuff working.

That honour would go to PHP, as far as I'm concerned. JS has many more fail-hard scenarios, and if you don't adapt your attitude to learn about things before using them, you're very quickly going to run into brick walls due to the entire callback thing.

Neither really has anything to do with MongoDB either - there's nothing inherently JS-y about MongoDB (JSON isn't JS), it just happens to have picked up a lot of traction around those parts.

To the comments that pointed out that JS and MongoDB aren't related, I'd argue they are, if only because the communities have related them.

The "MEAN" stack exists, MongoDB is popular in the JS community, and I feel they both exhibit many of the same hints of a fundamental lack of understanding about tools from a significant number of users.

As I said, there are great JS devs, and I'm sure there are some people using MongoDB for the 'right' things, but I don't think it's the majority.

Seems like you are saying Node is the new PHP or Ruby/Rails. It definitely has the most traction/gains so it has many experienced but also many newer devs.

Mongo 3.0 contains an engine that was acquired and is very decent, much like InnoDB in MySQL.

Having read https://gist.github.com/mapopa/3619146, I'm not eager to trust the judgment of a team who was ever willing to handle users' data so recklessly. Same reason I avoid MySQL when I can.

The authors of the new storage layer mentioned in the GP are the team that originally wrote BerkeleyDB, which has been used for misson critical workloads for decades. One of the authors actually invented Serializable Snapshot Isolation (SSI), the concurrency control mechanism used in Postgres. If you don't trust them to handle your data, I don't really know who you should trust.

For quite a long time, mongodb defaulted to buffering writes (presumably to win benchmarks) and dropping errors on the floor. Storage is not the only layer at which their dangerously poor judgment could screw me and my users.

At Couchbase we recently introduced an open source SQL-style language with extensions for nested data. Live tutorial: http://query.pub.couchbase.com/tutorial/

We get a lot of folks who start with a Mongo evaluation and end up moving to something more serious.

My only experience with DBs has been MS SQL, MySQL and PostgreSQL. I've always been intrigued with CouchDB. How friendly is CouchDB/PouchDB to someone from a SQL background?

CouchDB doesn't have a SQL like query language, and all acces is via http. Couchbase is higher performance with a binary protocol and autoscale features. So they are very different but they have in common that you store Json and then decide how to query it later.

Isn't MongoDb after 3.0 release quite OK? Especially when used with new db engine?

Also, mongo has very good drivers for every language, likely that is a good reason.

Ok then... Install your heavy SQL database on your super fast machine and use it for everything till your memory screams.

I'm not saying SQL databases are bad. It's just... They are not the only choice! You should be happy that now we have tons of alternatives to choose.

Sigh.

You know why many SQL databases use a lot of RAM? Because it improves their performance. They don't require the RAM, they just use it when it's available. Would you rather that it uses your free RAM to speed up your database operations, or to just have it sit idle doing nothing?

As for not being "the only choice" - while technically correct, this is very misleading. SQL databases and schemaless/document stores are _not interchangeable_. They are used for different usecases with different data models and different integrity requirements.

MongoDB is a special one entirely - even for a schemaless document store, it's just very poorly designed and maintained. Use any of the many other such databases - which gained a reputation by actually working well in production, rather than its developers organizing lots of hackathons as a marketing stunt.

1. SQLite isn't heavy. There are probably even lighter-weight relational solutions, but I haven't needed one, so I haven't looked.

2. Even among non-relational document storage solutions, Mongo has a poor reputation. As someone else says elsewhere in the comments, it's hard to trust a team that let so many ridiculously dangerous bugs (and a basically unreliable design) out into production for so long.

No SQLite isn't heavy but I hope you mean for use in development only.

The issue isn't MongoDB is bad and relational DBs are awesome. The real issue is new developers being taught that Mongo and NoSQL in general are the databases you should be using with Node. The problem secondary to that is devs not knowing when each is appropriate.

Sqlite can and is widely used in production server-side.

It's about time the "develoment only" myth was laid to rest. Please read the links bellow:

https://charlesleifer.com/blog/sqlite-small-fast-reliable-ch...

http://stackoverflow.com/questions/1711631/improve-insert-pe...

http://stackoverflow.com/questions/784173/what-are-the-perfo...

SQLite is the most widely deployed database engine out there.

Especially if you count all the phones and copy machines and AV equipment. Embedded databases are fun to code with.

Not in production it isn't.

It's actually commonly used in production for certain types of databases. SQLite is not a replacement for something such as postgresql but it certainly has it's use cases.

Do you mean for public facing sites? If so, what are the use cases?

The use case for SQLite isn't high traffic public facing webpages. Here is their own list of 'famous' users: https://www.sqlite.org/famous.html

Every iPhone, and Android phone have SQLite running on them, which easily makes it the most used production database. It's use case is typically very low traffic websites or as client side storage.

Maybe I'm missing something but is Mongo in use on the client side in that same way? If not then SQLlite is being shoehorned into the discussion and the context in which production was used should be clear. I feel like people are trying to be technically right instead of following the actual discussion. I never meant client side software and definitely didn't mean someone's micro traffic blog, where a toy database could be used.

It is interesting how much penetration SQLlite has on the client, though.

At any rate, I know better than to not be extremely specific so I brought this on myself.

It seems like there have just been some assumptions going both ways.

This all started when someone trashed RDBMS for being "heavy", implying that Mongo is good because it's light.

Why would you care how heavy a DB is if it's not on the client side? The kind of memory and storage you need for even the heaviest popular RDBMS (MySQL?) is still low enough that's nearly free to create a small app.

The only reason I brought it up is that there is a myth that many developers believe SQLlite is only useful as a dev enviroment database

On a multiuser system, that's a reasonable assumption.

That's partially true, but it does seem to find it's way in production on lots of low traffic sites.

My browser tell me something other.

I don't mean 'on the desktop' when referring to database use in production.

If you don't have the resources to run PostgreSQL, you sure as heck can't run MongoDB.

And if you really have minimal requirements, I recommend SQLite.

I'll just leave this here.

http://cryto.net/~joepie91/blog/2015/07/19/why-you-should-ne...

SQL databases are not heavy and d not require super fast machines.

There are options like RethinkDB, Firebase, etc.

Reply on news.ycombinator.com