SQLite hadn't been kicked to the curb, the WebSQL API whose “standard” was essentially “act like version x of SQLite” was kicked to the curb.
The concept of WebSQL was good, but the standard itself wasn't great, and the absence of any intent to implement from other vendors or intent to improve the standard from those who had embraced it made it a dead-end standard.
Nobody said what was so wrong about that.
I just don’t think there’s any rule that says that we need multiple implementations of thing to make it a standard. Also, we already have a standard for SQL and they could’ve just used the parts of SQLite that were compliant with that.
> I just don’t think there’s any rule that says that we need multiple implementations of thing to make it a standard
There are rules that say that we need multiple implementations of thing to make it a standard:
* The IETF requires it to be accepted as "Internet standard", ie the highest form: https://en.wikipedia.org/wiki/Internet_Standard#Standardizat...
* The W3C strongly desires that proposed standards have multiple, independent implementations to advance to the state of "Proposed reccomendation": https://www.w3.org/2005/10/Process-20051014/tr.html#cfr
> Also, we already have a standard for SQL and they could’ve just used the parts of SQLite that were compliant with that.
That would probably have been the best solution, I agree. And then people would have complained about bloat again, that browsers are now required to expose a SQL engine, that they are too big, etc...
Anyway, there are multiple implementations of SQLite itself. It’s been implemented in JavaScript and .NET and Java as well I believe.
I know right, there are in fact several (which may have been part of the problem). Most SQL database vendors tend to stray slightly (or not so slightly) from the SQL standard that they target; SQLite is no exception.
Just unfortunate that the browser vendors couldn't come together and build a client-side SQL based API rather than punting and implementing a very basic key/value store on top of which someone will eventually reimplement SQLite. It's like 5 or 10 years from now we'll have what we had 10+ years ago, such progress!
It's hard to say which reason was more compelling for Mozilla, but my gut feeling is that they were betting hard on NoSQL and came up with convenient technical justifications afterwards.
Are you thinking of WebSQL? I don't remember a SQLite interface client-side in any browser. I might be wrong, but I thought IndexedDB was originally implemented on top of SQLite in both Chrome and Firefox. I think Chrome moved to a different backing store, but that has no bearing on the API.
https://hacks.mozilla.org/2012/02/saving-images-and-files-in...
Well, yes and no. The whole problem with WebSQL was that it effectively meant using SQLite and never anything else that wasn't completely bug compatible.
So the whole problem with WebSQL was that it was, de facto, a web standard built around giving an SQLite interface client-side.
Yes, exactly, but Web SQL uses SQLite as the backing store. IndexedDB is its own thing, a standalone key/value store, nothing to do with SQLite.
Were you writing the same queries on the client and the server? Generally client and server are different moving parts of the same machine, so they don't really share a lot of code anyway. (Perhaps in some cases computations could optionally run on client or server depending on the circumstances, but security concerns limit the opportunities to do that.)
Yes, it's a pain to have to know two different technologies, but it doesn't seem like this often leads to literally needing to scrap code and rewrite it.
Well, yeah, that's the idea, isn't it? These days every language under the sun compiles to javascript, why wouldn't you want shared model, validation, queries, etc.?
> Yes, it's a pain to have to know two different technologies, but it doesn't seem like this often leads to literally needing to scrap code and rewrite it.
Currently working in an Angular 5 + TypeScript frontend with Scala on the backend, a shared nothing architecture, really blows compared to a single typed shared everything stack.
All the stats based queries run on the server? Throw 'em out and manually recalculate on the client, probably literally 100X the amount of code; less efficient and error prone to boot. Same for form validation, model layer (hey, why does this new property exist client-side and not on the server?), etc. -- it's a house of cards.
It's deeply disappointing to not be able to use SQL on both client and server. Yes, there's Lovefield, which employs its own DSL that has nothing to do with SQL, so you can scrap everything on the server and redo it on the client.
Or just give up and go with NoSQL on client and server, that seems to be what the browser vendors are implicitly pushing developers toward.