Adding Fuel To The Fire

This post may be a bit geeky for some people. If you do not use Twitter, have no idea what Twitter is, or could care less, then please skip this post.

There is a lot of discussion lately about the reliability of Twitter. As most Twitter users know, the service is up and down faster than a Smother’s Brother’s Yo-Yo. Wow, I really dated myself there, didn’t I? Ouch.

I am not going to address the up/down time, the technical knowledge of the staff, or the “transparency” buzzword everyone seems to be using this week. Everyone is quite aware that Twitter has its issues. Most of us hope that Twitter can overcome these issues. Most of us would have a huge void to fill if Twitter disappeared tomorrow.

With that said, I think it is unfair that some people are bashing the Twitter staff because they chose to remove a function of the Twitter API that was causing serious load issues on the server.

On May 25th, Alex Payne, one of the Twitter developers, made an inquiry on the Twitter development mailing list. He asked those who develop Twitter based applications if they were using a particular function.

One of the methods that hurts us the most is what we call the “profile with friends”. Via the API, you’d access it via this URL if you wanted the timeline of updates by bob’s friends:

http://twitter.com/statuses/friends_timeline/bob.xml

We’d still allow you to retrieve your friends_timeline for the authenticating user, just not the authenticating user looking at another user’s timeline. This latter method forces us to do a fairly complex query that, when not cached, can be pretty painful.

How many of you actually make use of this method? Should it stay or should it go? I can say that the equivalent view on the web is going away in all likelihood. You’ll still be able to see who another user follows, but not in timeline format.

Over the course of the next four days, more than a dozen application developers (including myself) responded that we were not using the function at all. Only two developers indicated they were using it, and both of them said while they wish it would stay, that they would be in favor of removing the function from the API.

On May 29th, Alex announced on the list that it was going to be removed, most likely the next day. I think the Twitter staff did an outstanding job of notifying developers that the issue needed to be resolved. Alex made it clear that this function call (whether from the website or the API) was causing issues, and needed to be removed. It was clear that they needed to remove it as soon as possible. Did they wait long enough? Did they wait too long? Did they hurry too fast? If I had a service that was being brought down by a specific function call, I don’t think I personally would have waited four days to remove the problem code. I commend the Twitter staff for being patient (as well as diligent) while waiting for responses from the application developers.

It’s been two days since the function was removed, and 5 developers have responded that their applications are now broken because they were using the function. Some of them feel that the Twitter staff did not provide enough notification to the developers (even though they posed the question on the development mailing list). They feel that they should have given developers more time to discuss the issue (the function was breaking Twitter, what needs to be discussed?).

It sucks that the change broke some applications but I think it would suck more if the Twitter staff ignored the very things that were helping to bring down the service in the first place. I have noticed that even though the function is gone, Twitter is still up and down lately, so it’s clear that the function in question was not the sole problem, but then again, I don’t think anyone thought it was.

It should be noted for the non-geeky people who made it this far, that the function they removed allowed you to see a users timeline with their friends updates included in the timeline. If you used the Twitter website, you will remember it (it was also removed from the website). It wasn’t a crucial feature, as indicated by the number of developers who were not using it in their applications, but some are calling for the Twitter staff to bring back the function.

It should also be noted, that most of the applications that used that function did so to get around the rate limits on authenticated API calls.

I apologize for the long winded post. I know everyone is concerned about the fate of Twitter, but I really don’t think the type of bashing I have seen today is called for. It’s not productive for the Twitter staff, or any of the developers who want to see Twitter succeed. I also don’t think the removal of this one function, that apparently caused a lot of issues for the Twitter servers, means their API has “degraded”. It only shows that the Twitter staff is doing everything they can to ensure they bring stability to the servers, and for that, they should be commended.

– Powered By Stuffr! –

Technorati Tags: Twitter, development, API, issues
Sphere: Related Content

Posted on June 1, 2008
916 words · print

1 Comment so far

1

We Don’t Know What We’re Doing But We’ve Got the Money to Fix It| Zoli’s Blog on June 13, 2008 comments:

[...] Adding Fuel To The Fire [via Zemanta] [...]

Leave a reply

Name (required)

Email (required)

Website

Comments

If this is your first time commenting on Slobokan's Site O' Schtuff, your comment will be held for moderation. Once your comment is approved by Slobokan, any future comments will not be held for moderation. Comments may be edited or deleted at the discretion of Slobokan, and may be re-produced at any time for the purpose of discussion, argument, or ridicule. This policy is not open for debate. If you do not agree, do not comment. It's that simple.