Three Monkeys, Three Typewriters, Two Days

October 28, 2005

Iran controversy reactions

I've been looking at the reactions to the articles about the statements the president of Iran made regarding maps of the Middle East. This is things like user comments on the BBC site, etc. Meta-reactions to his statement, so to say. Some interesting conclusions, very few of which are correct, could be drawn from the content of these comments. These jumped out at me as representing the uninformed consensus of the world:

  1. Iran is an Arab country.
  2. A nuclear weapons program is illegal under international law.
  3. In 1948, Israel took away all the land the U.N. had decided was for the Palestinian state, leading to large numbers of refugees and our present situation.

I just wish people would do a little more reading for item 1. The next step, for item 2, is to realize that not all countries are signatories to the Nuclear Non-Proliferation Treaty (and that in fact there are at least two current declared nuclear powers who are not signatories). For non-signatory states, there is nothing inherently illegal about pursuing nuclear weapons capability. Finally, for item 3, I suggest people look at a map of the Middle East in 1949 and compare it to a map of the proposed U.N. partition of Palestine. That would make it abundantly clear who took over most of the land for the proposed Arab state in Palestine.

Posted by bzbarsky at 3:33 PM

October 26, 2005

Gecko 1.8rc1

We're now locked down for Gecko 1.8rc1 -- the first release candidate of Gecko 1.8. I'd like to urge everyone to download this (in a Camino or Seamonkey branch nightly, or in the Firefox 1.5 release candidate once that's available) and bang on it. Camino testing is especially welcome, since Camino exercises a lot of important embedding code that's not exercised by either Seamonkey or Firefox, but is used by a variety of other apps. If there are stop-ship bugs, we really need to know, and the sooner the better.

I don't really think we'll find many stop-ship issues, though. This is by far the best release candidate we've ever done. It's actually a real release candidate, unlike the previous ones!

Posted by bzbarsky at 9:39 AM

October 21, 2005

Reviews

I started reading through my request queue today. These are requests for review I've made from others. It's pretty depressing reading, as usual. About half of the requests are on behalf of someone else, not even for my patches. Fully half of the patches date to June 2005 or earlier. Most of these have no indication from the reviewer that they ever plan to look at the patch. And this is given that I go through this list every so often and harrass people with non-bugmail... Maybe we should just start publically naming names. A "review hall of shame" if you will, posted on Mozillazine every week. Note that I fully expect I would be in any such list, since I have at least 4 or 5 review requests from before September 2005... So I understand how this sort of thing happens. But at least I try to make an effort to tell people in the bug if review will take a while...

Posted by bzbarsky at 10:03 AM

Expired bugs -- call for help with reopening the ones with patches

The unco bug expiration seems to have indiscriminately expired bugs that had patches attached (eg bug 287388), though I'm pretty sure this was brought up as an issue. While I fail to believe that this could not have been avoided, the damage is already done. So what we need to do now is to find all expired bugs with attached patches and reopen them... Then probably make sure that the patches have reasonable reviews requested for them.

Unfortunately, I'm a little swamped with this whole Gecko 1.8 thing at the moment, so I can't really go through and do it myself. Could someone do it as a favor to me? Please?

Of course the entire review situation for people who post a patch to a bug is awful. A lot of these patches never even end up with a review request, from what I can tell. We need to be a lot better about welcoming contributions; perhaps weekly scans for attached, non-obsoleted patches without review?+- flags? Followed by requesting of review as needed? If bugzilla can't do this for us, we need to fix bugzilla. I realize that the first time we do such a scan we'll hit a lot of false positives, but we should just go ahead and obsolete them. We do it once, and we don't lose track of patches thereafter.

Posted by bzbarsky at 9:54 AM

October 14, 2005

Drowning in bugspam

If you cced me on a bug sometime in the last two days, I probably won't get to that mail until well into the middle of next week, given the current volume of bugspam that this latest round of UNCO-twiddling has caused... If it's urgent (as in, affects 1.8), please just mail me directly. Review requests and review/approval being granted are probably OK as they are, since I can filter for those.

That said, so far, this round seems to be less noisy than the first one. Of the first 7 bugs I looked at from the list of newly-EXPIRED bugs I got mail for, only two were possibly valid bugs (one waiting on me to decide whether it's valid, actually, one just in need of deciding on the component before it's confirmed and then fixed). The ratio was much worse with the first round.

Posted by bzbarsky at 1:43 AM

October 9, 2005

Java plugin fun

I recently installed the Sun Java plug-in for the first time in years to test a patch I was making. A bit of testing quickly showed that:

  1. It has a tendency to randomly hang (and hang the browser with it, of course).
  2. It has a tendency to assume that any failure on the browser's part to respond to some sort of IPC-type stuff (e.g. because the browser is suspended in a debugger) means that it's time to print out that there is a bug in the browser and call exit().
  3. It just happens to leak both the plugin host and the JVM manager.

Combined with the code I was actually changing, I devoutly hope I have as little interaction as possible with plug-ins in the future.

Posted by bzbarsky at 3:03 PM

October 7, 2005

Profiling of script access to the DOM

I spent some time today profiling some "access lots of DOM objects" testcases. Filed a few bugs, but even with those accounted for we're slow. Probably 20-30% slower than Opera 8 (without the bugs being addressed it's a lot worse). Worse yet, there's nothing obvious to point at; just death by thousands of paper cuts. Here's hoping that we can at least reduce the amount of locking and such XPConnect does so that the common case (DOM, UI thread) will be a little faster...

Posted by bzbarsky at 8:54 PM

October 4, 2005

Shana Tovah to everyone

May you have a happy year 5766.
Posted by bzbarsky at 1:30 PM