Evan Williams: Twitter not abandoning Rails | Hacker News Reader

Evan Williams: Twitter not abandoning Rails

twitter.com

53 pointsnickb41 comments

Now where's the TC retraction? (oh, forgot, there's no link-bait in boring stuff like accuracy)

retraction or correction? TC is only partially right/wrong... Ev's says:

"Lots of our code is not in RoR, already, though"

Sounds like an ongoing process rather than a new thing

Twitter has been mostly not Rails for a very long time. You can't build a service like Twitter on top of a traditional relational database stack, so much of their infrastructure is based around custom message queue stuff.

Let me give you a scoop, then: Lots of 37signals code is not in Rails, either. We run Solr for search. We use a C-based poller for Campfire. We call out to ImageMagick for transformations of images.

Most successful LAMP sites use software from more than one bucket. That is not news.

Oh, there may yet be.

Melodramatic mea culpas can attract traffic too. Getting a story wrong can thus provide a two-fer: one batch of attention for the sloppy, inflammatory first piece, then another for the contrite (or alternatively, defiantly unapologetic) correction.

Ever see the movie 'The Corporation', which diagnosed the 'person' of a corporation as being psychopathic? Well, the DSM-IV diagnosis for an ad-supported blogging site will tend toward Histrionic Personality Disorder†. Clickonomics almost demands as much.

† http://en.wikipedia.org/wiki/Histrionic_personality_disorder

unless it's a major mistake, most publications don't really publicize corrections (especially not with blogs being able to edit their content.) UPDATE

Indeed: the story (http://www.techcrunch.com/2008/05/01/twitter-said-to-be-aban...) now has an update:

"Update: Regarding Evan Williams’ statement here [link to http://twitter.com/ev/statuses/801530348], all I can say is that multiple sources claim that Twitter is telling people they are planning on moving away from Ruby on Rails. This is not the first time a company has denied something that has turned out to be 100% true."

Looks like they chose the "defiantly unapologetic" correction. I wonder, have they done any studies on which type of correction gives them more traffic? Might be worth it.

There's an important point that is getting missed here.

Twitter uses multiple technologies, not just Rails.

Everyone who hopes that their site grows to a Twitteresque-size should design a system such that components (db, messaging, processing, frontend) are decoupled enough that they can be swapped (ideally transparently) if/when they become the bottleneck.

More details on that statement:

"Selvitelle told eWEEK that reports of Twitter abandoning Rails are "Not true in any sense. We use Ruby as our primary language. We have plenty of back-end architecture in other languages. Especially prototypes. We still use Rails and have no plans to discontinue this in the future."

http://www.eweek.com/c/a/Application-Development/Ruby-Rails-...

Twitter could have been such a nice ActiveMQ/Jetty+Cometd Java application I think. And probably run on half the amount of hardware that they have now. Rails is not very efficient when it comes to resources. I have re-implemented Ruby/Rails code that runs faster and in 3 times less hardware on Java. Closer to the metal. Better usage of OS services.

It's a 500 error right now. sigh

edit: ok, it's back.

Yep, got a 500 error to! I thought the post was a joke...reloaded and realized it wasn't, got a clean page.

If you reload the page 3-5 times in a few seconds, you get the 500 error page. Now those are serious scaling issues!

Afraid so. Rails got them to market, if they have to port to whatever, they should do it and I don't see why that is a big deal for anyone. I still like rails even if it can't scale to twitter.

WOW - you are right. I am seeing that right now. Thats terrible. They could do a server too busy message and it would be more correct I think.

They lost a DB server:

http://twitter.com/twitter_status

Recent downtime (the 500 errors) caused by rebooting database servers...

I just did a double-take... rebooting servers... really? really??

http://blog.twitter.com/

Is this the first time a twit has been submitted to news.yc?

Nope. Cringely gets submitted here with some regularity, for example.

Say what you will about Cringely, he's had pretty stunning accuracy at predicting Apple's biggest moves for several years now...even the stuff that seems unlikely except in hindsight.

Yeah, I just couldn't resist the snark and he was the first soft target that came to mind :-)

And, I don't know if you've heard Joe Kraus speak, but he almost always tells the story of the serendipity that led to his success with Excite...and the first step in that series of unlikely events, the kind fellow who spoke to him was Cringely. He lined up their first customer and first investor and offered them lots of (bad) advice. But Joe obviously still feels very positive about it, because he's brought it up three times out of four that I've seen him speak.

Cringely's telling of the story:

http://www.pbs.org/cringely/pulpit/1999/pulpit_19990121_0005...

Are they already using Java?

This is al3x's comment from last night:

"Apologies, that would be our Apache answering when it can't find an application server to talk to. We'll work returning a proper 501 in that case. "

http://groups.google.com/group/twitter-development-talk/brow...

It's been a few years since I did Rails, but isn't "app server" the parlance of the Java world?

Application server is not a Java specific term. In the case of Rails apps deployed with Mongrel the mongrel instances are the application servers.

sigh

I admit, this is really hard to resist.

But please, can we try?

"We lost a database server, which is causing lots of errors. Reloading the page may work. It's in the process of recovery."

From http://twitter.com/twitter_status

Maybe they can use their new funding to go hire some engineers who know what they are doing.

I'll get karma-bombed into oblivion for saying this, but I've found it disturbing that a programming culture so obsessed with "testing" has so many problems keeping their servers up and running...

That's because Twitter's architecture was designed using the Bob Dole school of infrastructure...it can't stay up.

Even Hacker News seems to have skipped making load testing part of the process. Unit testing won't help you with these types of problems, but load testing is a pain and big companies skip it too sometimes, usually to their detriment.

http://blog.twitter.com/2008/04/welcome-john-and-steve.html

They're off to a great start...

That should be a really fun project, I wish I had those kinds of problems to solve.

"You just can't buy that sort of entertainment" - https://financialcryptography.com/mt/archives/000959.html

OR twitter abandoned rails long ago... nothing to see here

Still 500.

And to think, I nearly based my company on Rails... Not a great advert for what appears on the face of it to be such a promising Framework.

They're having problems with their database servers. It has nothing to do with Rails.

I love the mentality here. Thrash TC for posting something they "know nothing about" yet turn the attention onto something else that you probably don't know anything about. Unless you guys have access to Twitters racks, let's save some of the defense/speculation. I'd guess Arrington and Co. probably have significantly better sources than anyone on here.

They announced that it was a DB problem yesterday.

http://twitter.com/twitter_status

If they said it, it must be true!

If your company needs to scale to the level of twitter, using rails will be a nice problem to have (it would have got you that far quickly).

Reply on news.ycombinator.com