Chris Anthony
(I think disqus limits the depth of comment threads, which is unfortunate but understandable.)
I did see the update - in fact, I paraphrased part of it. ;) But I think we're talking past each other.
I understand that you're frustrated with Twitter's stability. I am too! I want Twitter to be there whenever I want it, and I want Twitter's tech support guys to have nothing to do but sit around and tweet. Sadly, that isn't the case. But a boycott isn't going to change that, and honestly it's not even going to start to change that. Like I said above, it's largely an exercise in making the participants feel better.
I agree with you that Twitter needs a fallback server, even if it's just a Windows NT 4.01 box that displays a filing cabinet with a poster saying "Beware of the Leopard" on the side. Part of the problem, unfortunately, is that Twitter is tossing up the "There is something technically wrong" pages with enough frequency that they're going to have to make a fallback page very significantly different - so much so that the average user will actually read it, rather than just assuming that it's a run-of-the-mill error and keep refreshing. And then, once they've got the fallback in place, the mechanism to do the falling back needs to be incredibly fine-tuned - how long does it wait for a response from the main server before it kicks in? If there's non-server latency, how often will it throw false errors? It's not an easy problem to solve, especially for a company that's growing as fast as Twitter. That isn't to say that I think they should give up - just that it's a hard problem and I don't think Ev and the group really expected this much growth this quickly.
Back to the main point: I agree with you that the conversation is great, and I'm honored and flattered to have been able to be part of it. But the thrust of the conversation is, at least nominally, the pros and cons of the Twit-Out - and so far, at least for me, the pros don't outweigh the cons.
If you're willing to entertain suggestions, it occurs to me that a better use of May 21 might be gathering a group of talented developers and designers, and creating your own fall-back server! Set up an infrastructure like Twitter's, active only when Twitter is down (so as not to compete with Twitter), and storing tweets for users; the system could then, when it got a reliable response from Twitter, forward all of the stored tweets on (so that they were placed in Twitter's database), and then close itself down again. Call it the Emergency Tweetcast Network or something like, and make sure people know about it as a backup for when Twitter's servers go on the fritz again.
(If you're lucky, Twitter might spend some of that $75mm in VC money to buy your ETN... ;)
Chris Anthony
Ah, reply depth limiters.
I can see that argument. On the other hand, they haven't yet: and while we'd like to see Twitter be the Perfect Application, we also need a way to deal with the downtime until it is the perfect application. ;)
Another thought: if you must boycott, could you rig a script to tweet "Twitter, please upgrade your infrastructure to avoid future downtime!" or something like that, every half hour or so? That way you're getting the message across without a bunch of people nudging you. ;)
sreiser
The inability to update was my fault. I set disqus with a thread depth of 5 which is generally enough for my content, sorry.
As far as how they treat folks when it comes to notifying us about outages, I can understand and respect that as a point. (Although when the had the twitter is down cat, it made me hate cats). If this was posed as a "We're boycotting because we feel the folks at twitter don't keep us in the loop when there are outages" I'd understand and respect it. This feels more like "I'm pissed that twitter goes down so I'm taking my ball and locking it up for a day.. that'll teach them".
Chris Anthony
Holy crap, that was long-winded of me. I'm so sorry. >_<
Chris Anthony
How do you know that they aren't tested before they're deployed? To be blunt, I have run into many, many instances of things running fine on the local server and crashing the live server. One variable set in a slightly different way can spell disaster.
Even then, how is the Twit-out going to get the message across? You're trumpeting "community is key" in the actual Twit-out post; you don't talk anywhere there about how server fixes should be tested before they're implemented. Sure, that's a component of maintaining a stable server environment, but that's an awfully broad topic, and - again, to be blunt - nobody but the Twitter developers actually knows what their server environment is like. If they are testing before implementing, and what they actually need is better servers or server software, then saying "you have to test before you implement!" isn't going to do anything. If they test and have great servers but it's Ruby on Rails that's failing them as they scale up, then nothing is really going to help because they're going to have to rewrite the whole damn thing if they want to get it anything like stable, and that's going to take months at the least.
In my job as a medical data analyst, I'm often confronted with the problem of separating the immediate cause of an injury from the underlying cause. A person may have broken their leg because the leg hit a table, but they hit the table because they fell, and they fell because they tripped. Tripping is the underlying cause; hitting the leg is the immediate cause. In the current case, a server upgrade that deleted server caches and brought the server to a halt while the caches were rebuilt is the immediate cause of instability - but we don't know the underlying cause, and we can't unless we're actually in the Twitter server room or talking to Alex Payne, who's their main tech guy.
Like I've said elsewhere, I'm nothing but sympathetic, and I want Twitter to be stable as much as the next guy. I just don't think a Twit-out is the best way to the goal.
