Let's say they changed the architecture so that @replies go to a dedicated queue, and not to the public timeline queue. The solution is simple, then... make two copies when posting, one for each destination. But I don't think this is the issue, since supposedly mutual followers will see the @reply in your public timeline. So I really don't get it.

Twitter--. They are either incompetent, or are lying.

The scaling issue was technically the second excuse. They initially said it was a user experience issue.

But I really liked the part of your post where you were like, "assume they have queues like this, and then all you have to do is that!" Because I'm sure you're right, and to a competent developer, everything at Twitter is that simple.

I am however a bit confused about the "lying" part. Uh... why? What would be motivating them to decieve you?

Each paragraph was an independent thought.

  A)  They are lying.
  B)  I wonder why they lied. I guess it might be hard if...
  C)  Ok, they might just be stupid.

The "gigabytes of data" thing doesn't mean anything on its own; you have to consider what you're doing with the data, and what the operating constraints are. I don't think this argument is very compelling. For instance, could you retrofit the ircd source code (leaving everything in IRC protocol) to handle Twitter's work set?

I'm sure Twitter could be optimized quite a bit from where it is today, but it isn't one Rails neophyte's side project anymore; there are very smart people with lots of resources working on it, and I'm inclined to believe those people when they say parts of the problem are Hard. Haven't you ever been surprised by how Hard a simple-looking problem turned out to be?

I find computing problems tend to become hard based on the constraints not the bassic problem. A friend of mine needed to create a simple counting program that sent the results back over the network. The problem was he needed to do it in less than 340 bytes of ram, on a 8khz chip with 2kbites of permanate storage. AND he needed to validate the memory so the program could sit on a chip in the field for 10 years without problems even as the memory got randomly fliped.

That's a hard problem that took a few months but hey it works just fine. I expect Twitter to have a few anoying problems to track down, but they are long past the point where they have no budget and no time to solve them. Worst case they change some of their constraints and move on.

PS: My problem is not that twitter is trying to fix somthing dificult and limping by my problem is they have been doing it for as long as they have.

They are still running the regexp to replace @twittername with a link anyway (because it works if the @username appears later in the tweet).

http://blog.twitter.com/2009/05/we-learned-lot.html

The jist: they're bringing back the old behavior only when the message starts with @username (and hasn't been created clicking the reply icon).

It sounds like they made a technical change which trumps the actual message content with a piece of reply-to metadata, which is either explicitly included by the client or parsed by their infrastructure.

Previously I believe it was "dumb" and didn't examine anything, and it was up to the client to filter based on the content of the message when it came to showing half-conversations.

Good thing Twitter bought Summize, as it's quickly becoming the most useful part of Twitter. Search "to:username" gives you this exact functionality.

  Your username: you
  You are following: friend
  You are not following: stranger.

friend tweets: @stranger I wish the would be like Clojure or better work on an Iterator instead of lists

The way it worked before, you would have the option to see both of these messages, one of these, or none of these. Now you get to see none of them. @you tweets are unaffected.

I'm finding it hard to find a clear description of this anywhere, so I might be incorrect.

Personally, I liked seeing conversation fragments. If the reply looked interesting I could easily enough click on the other person's username to load their tweets and get the question/conversation.

The ironic thing about Twitter forcing this is that those of us who wanted it this way are going to be screwed, too. People are already prefixing @replies to get around the issue. The preference has moved from the receiver to the sender. If someone wants you to see a conversational reply they'll just do "PR @username" or whatever.

The result is that has Twitter actually removed all choice for people on both sides of the issue. Whoops!