A lesson in optionality

[jducoeur]

[Side-note: I haven't been posting much. Sorry: work's been eating my brain, and I don't want to bore people with a complete excess of CommYou stuff.]

Facebook's latest feature is built-in chat: when you're in FB, there is simply chat *there*. It's not something you turn on -- it just sits at the bottom of your window. And you Can't Turn It Off. (Far as I can tell.)

Which wouldn't be so bad, except that it's taking up about 60% of my CPU, far as I can tell. I was initially worried that something in CommYou was running out of control, but no -- it's just as bad on the Facebook homepage. And since Chat is their new feature, I rather suspect it's the culprit.

Moral of the story: features that the user can't turn off are *bad*; cute features that they can't turn off are worse. A lesson for me to remember in CommYou...

Comments

[roozle.livejournal.com]

That ... explains a lot of badness that has been happening to me with computers lately.

thank you for posting that.

[sephiaowl.livejournal.com]

Actually, you can just mark yourself as unavailable, and I *think* that pretty much turns it off.

[jducoeur]

Tried that -- it doesn't seem to help. I get the impression that that just changes the way you present your own presence, but it's still doing all the work to fetch *other* peoples' presence, which seems likely to be costing all the CPU. (At least, that's my guess...)

[serakit.livejournal.com]

Okay, now I'm definetely not getting a facebook until they fix that. I eat enough of my CPU as it is. (Once they've fixed it, I'll come play with CommYou.)

And two or three posts every couple of days is "not much"?

Huh

[metageek.livejournal.com]

I've still got 3 Facebook tabs open from 2 days ago, and Firefox's CPU usage is minimal (1-2%). Maybe their chat stuff just doesn't work on Firefox for Linux?

Re: Huh

[laurion]

No problem here. Nor on the Mac.

Re: Huh

[alexx-kay.livejournal.com]

Wacky hypothesis: Something about Jducouer's machine environment flags it to the chat code as "this is a development machine, so I should run in debug mode".

Re: Huh

[jducoeur]

Conceivable, although I don't know why it would get so flagged. Remotely possible that the presence of the Developer app might do that, I suppose, but it would be rather odd...

Re: Huh

[jducoeur]

Curious -- this isn't Linux, it's Windows.

Now you have me wondering, though. Maybe it's something strangely browser-session based. The CPU problem is very consistent, but I haven't restarted the browser in a good while (most of a week). Maybe it's a pathological situation that doesn't arise for quite a while, or is related to navigating away and back many times, or something like that...

Re: Huh

[metageek.livejournal.com]

Curious -- this isn't Linux, it's Windows.

Right. I meant that maybe the reason I wasn't seeing it eat CPU was that it just wasn't working for me.

I haven't restarted the browser in a good while (most of a week).

That could be it, then. If it's FF2, that could definitely be it; I try to keep mine open indefinitely, but I always have to restart sooner or later, what with memory leaks and all. (3.0 is supposed to have fixed most of those.)

Actually, memory leaks could explain the CPU usage: garbage collection can be extremely CPU-intensive in pathological cases—e.g., when JavaScript is repeatedly polling the chat server, constructing and discarding objects representing the connections.

Re: Huh

[jducoeur]

If it's FF2, that could definitely be it; I try to keep mine open indefinitely, but I always have to restart sooner or later, what with memory leaks and all. (3.0 is supposed to have fixed most of those.)

Yaas -- probably the biggest reason I'm looking forward to 3.0. I run a lot of Google apps (email, calendar, reader, etc), so my memory usage tends to be pretty bad. I've been controlling that lately by not running quite as much at once (three windows at a time with half-a-dozen tabs total, instead of seven windows with 20-some), but I do wonder if the cumulative memory leakage is somehow interacting badly with Facebook.

I still suspect Facebook Chat is complicit (it's the only thing on the FB homepage that *should* be using a non-trivial amount of CPU); as you say, that sort of communication can be expensive, especially if the implementation is a little naive. Now I'm wondering what their protocol is, and what I can learn from it. CommYou is going to have some automatic polling, at least by default, and it would be good to not have similar problems...

Re: Huh

[metageek.livejournal.com]

Now I'm wondering what their protocol is, and what I can learn from it.

If it's in JavaScript, the source is sitting in your cache.

You could deliver something in Flash or Java that maintains a persistent connection, so that you can push stuff down to the client. 'Course, that means a lot of connections on your end, which gets expensive.

Re: Huh

[jducoeur]

If it's in JavaScript, the source is sitting in your cache.

True, although we'll see if that's useful. It is quite possible to obscure Javascript to a fantastical degree -- some of the better compression programs produce output that could win an obfuscated-code contest in a walk. (At the highest compression levels, the results have left me scratching my head at the idea that it could even execute.)

You could deliver something in Flash or Java that maintains a persistent connection, so that you can push stuff down to the client. 'Course, that means a lot of connections on your end, which gets expensive.

Yeah, that's not likely to be practical. Current plans are for variable-speed polling, sensitive to the context and nature of the conversation...

[laurion]

Huh. I'm not seeing any problems. I actually like the way they've implemented it; comparatively unobtrusive, and the notifications popup down there is also in the category of useful to me. I play Chess and Scrabulous, and it alerts me to when it is my turn. Unlike many web based chat systems, it eats a minimal amount of real estate when not actively engaged.

I agree that turning it off should be an option though.

Amusingly, Digsby (whih I can no longer use) has _already_ incorporated FB Chat.

[jducoeur]

Yes, and I'm rather curious about how they did that -- far as I know, FB Chat isn't an open protocol. I can only assume that they reverse-engineered the traffic, and are poking into it. (Which, to be fair, is what Convoq did for *all* of the protocols, so I suppose I shouldn't be so surprised.)

I'm quite intrigued, actually. IM/Chat integration is in the cards for CommYou in the moderately near future, and I'm starting to wonder if there is useful synergy to be mined with Digsby.

Why can't you use it any more?

[laurion]

Doesn't run under my new Ubuntu install. At least, not until they release their intended linux version.