Older blog entries for dwmw2 (starting at number 133)

Bloody telcos.

For a long time, there was a second mobile phone on my account, which was no longer used (for various reasons) but never got cancelled. I paid Vodafone a fair amount of money because I never quite got round to cancelling it.

They have a duty of care to their customers -- arguably they should have contacted me and asked if I realised that I was still paying for this phone which was never even switched on. But they didn't.

Last year, I finally got round to calling them and telling them that I wanted to close the account.

A month or so later, I closed the direct debit and informed them that I had done so.

Now, I'm receiving threatening letters from a debt collection agency, demanding payment for the three or four months more which elapsed before Vodafone finally did close the account at their side.

Now, are they just being muppets, as we have come to expect from telcos, or are they deliberately behaving dishonestly in an attempt to defraud me?

The extra few months' line rental is nothing in comparison with what I paid during all the time I left it due to my own incompetence -- there's a school of thought that says I should just pay it and make them go away. But obviously I'm far too obstinate to let them get away with that.

Jack asks "why did we ever abandon Mutt and Pine?"

I don't know about you, but I didn't. I still swear by pine -- I use it on handheld devices a lot, because it does actually integrate well enough into that environment. With a touch screen, and pine's 'mouse-in-xterm' option, you can just tap on menu options, on mails in the folder index, etc. You don't actually need to use the fake on-screen keyboard for anything much in order to read mail with pine.

I also often find myself running pine on 'real' computers -- I often fire it up to read my mail while I'm waiting for Evolution, in fact. But the main reasons I don't use it all the time are twofold:

Firstly it's because I just don't think I could live without the little tree of folders which you see on the left-hand side of just about every GUI mailer these days -- the one which tells me how many unread messages are in each folder. That's integral to how I deal with mail these days.

Secondly, I very much like composing new email in separate windows, rather than in the main mail reader window. I have habit of hitting 'reply', perhaps making a half-hearted attempt to respond to an email, and then getting distracted and leaving it for days before I eventually find it the composer window on my desktop again, finish it off and send it. If I didn't do that, then stuff would just get lost and I'd never reply to it. I'm bad enough already.

For short-term use though, such as when I quickly connect by GPRS to check for new email in my inbox, pine is just great.

I'm not sure why Pete seems to suggest that, as a text mode application, cut and paste shouldn't work. Yes, they're not "integrated with the desktop" and they don't have dbus notifications or panel applets, but cut and paste generally work fine in an xterm. In fact, it's Evolution where cut and paste is generally screwed, in my experience. I've never had a problem using cut and paste with pine.

Hopefully the Sylpheed crash will get looked at more quickly than Evolution bugs do, and I'll be able to properly assess Sylpheed's suitability some time soon.

26 Mar 2006 (updated 26 Mar 2006 at 13:49 UTC) »

I'm not entirely sure why David is so happy about Jesse filing Evolution bugs. I've been doing that for years and it never really seems to have much of an effect -- even when I include patches, in some cases.

But since it makes him happy, and since I like making people happy, I suppose I might as well transfer a couple of the real showstopper bugs which prevent me from upgrading to FC5 from Red Hat bugzilla to GNOME bugzilla to see what happens.

Here we go then:

Bug #336074 (RH #167805): Check for new mail only in active folders, not in multiple years of archive folders.

And the real killer:

Bug #336076 (RH #183219): Evolution no longer uses STATUS to check for new mail -- instead it SELECTs every single folder in the system, downloads all headers and flags for every mail.

David Nielsen writes:

"Anyways, both Jesse Keating and David Woodhouse have been posting lists of Evolution shortcomings and one thing that struck me was.. where are the bugreports?"

For historical reasons, most of mine are here in Red Hat bugzilla rather than here in GNOME bugzilla. This is partly because for a very long time Evolution bugs were in Ximian bugzilla, which could never seem to be able to set and use cookies correctly, so I couldn't actually use it -- it would just ask me to log in over and over again.

Also, the Ximian monkeys were just too much of a pain to deal with. For example: even a simple RFC-compliance bug fix like the "don't use underscore in HELO" one, where I supplied a simple patch, was something which required me to argue with idiots. I believe the patch for that is still only carried in the Red Hat package rather than being accepted upstream.

Thankfully, there seems to have been a change of guard in Evolution-land in recent times. The new maintainers seem to be a whole lot saner, so things might be looking up.

"I'll make myself guilty of complaining over Evolution without filing bugs as well."

I try to make sure it's all in bugzilla but I've certainly been guilty of complaining without filing bugs occasionally, it's true. To a certain extent, I don't see the point in filing more bugs, while the bugs in NEW state about "Evolution crashes when I view the attached mail" are still being ignored. Once we actually start to look at the potential security holes we knowingly shipped in FC-5, I might be more inclined to spend some time making sure everything else is in bugzilla too. Until then, it doesn't seem worth it.

Now that I can actually use the bugzilla which we're supposed to use for Evolution, and now that the Evolution maintainers seem to have a little bit more of a clue, I suppose I should make sure more of these are in GNOME bugzilla and also attempt to clean up and submit some of the patches I've been using in my own builds for years... especially as I think I've given up on the idea of switching to kmail right now. Maybe I'll try it again at some point, but the weird way it shuffles the folder index in front of my eyes while I'm using it is just too screwed up. That needs to be fixed before kmail is a serious contender.

Having also been evaluating new mailers over the last week or two, I've a few comments to add to jkeating's mailer comparison:

Evolution supports IMAP over SSH, unlike Thunderbird and Kmail -- it can get at its mail server by running 'ssh $MAILSERVER exec imapd'. That's quite important for me, since the mail 'servers' I use don't actually have an IMAP dæmon listening, and the only way in is SSH. And I don't want my MUA knowing my passwords (or asking me for them either). Authentication is what I keep ssh-agent for.

Evolution certainly used to have an option for not displaying HTML mail. If it's not there in 2.6 then it's a regression. That's not a feature I'd want to live without.

The other thing that I found really suboptimal about kmail when I tried it was the way it handled deletion of messages. When you mark a message for deletion in any other MUA, it stays in the mailbox index. You can read your way down the mails in the folder index, deleting them as you go, and occasionally maybe glancing at the actual text of the mail. But kmail does something really weird -- it first crosses through the message in the inbox as you'd expect, but then it actually removes it. So as your eye is scanning down the folder index, kmail is moving it all around and making it hard to follow. Since I spend quite a lot of time just glancing at mail and deleting it, that's really quite an important misfeature for me.

Evolution can only check mail in all folders, or only the INBOX. That's fairly crap when 'all folders' contains many years of archived mail which is never going to change. What I want is to check mail only in active folders. I have patches to Evolution to do that, and kmail is capable of it too. It can't use the subscribed status of the folders, but you can configure each folder to either be included in the mail checks or not -- and it's not difficult to do that en masse by editing kmailrc directly.

Another thing I want from a mailer is the ability to send patches inline without it eating them. I know that Evolution can do it if you load them from a text file with the 'Preformat' style. I assume that Kmail can, but I'm told that Thunderbird will eat them.

Also, I want my MUA to be able to handle the standard way of sending stuff like party invites to multiple people:

To: Some people : ;
Bcc: my@friend.com, my@otherfriend.org

Evolution silently drops the To: header, while Kmail tries to qualify it by adding @infradead.org to it, and then complains that ':' isn't a valid character in it. Stupidly.

I also found sylpheed to be a non-starter, because it would segfault as soon as it connected to the mail server.

Pete, I'm not sure you count, since you've had access to the final FC5 tree for almost a week now anyway -- if you're only downloading it today, then you blatantly aren't one of the people who's desperate to have it while it's hot. But your answer does confuse me a little, in two ways.

Firstly: I've reported a bunch of installer bugs in recent times, and I don't remember Dave or Jeremy once suggesting that the problem is due to not using 'magic' ISO images -- despite the fact that I almost always use network installs because I think that using CD/DVD media is just a pointless waste of beermats. I appreciate that some people do have reasons for actually burning it to those round shiny things, but I'm surprised at how many take that option, and I'm vaguely surprised that you're amongst their number. Network installs are perfectly well supported; you seem to be implying that they're not.

Secondly: Even if for whatever reason you did want to use the 'magic' ISO images and deal with the unreliability that comes with using real CD/DVD media, they're really not hard to fetch -- you certainly don't need to muck about with anaconda scripts. For a start, the images/ directory is already there for download anyway.

I'm not sure if BitTorrent is as clever with its 'seed' as rsync is, but if you'd downloaded rawhide during the last week or so you could just 'cat' the whole set of RPMS and other files into appropriately-named ISO files and then rsync the real images using those as seed. You still end up downloading a few megabytes, but that's still a whole lot less than the 3.4 gigabytes you'd be downloading otherwise, if you started without a seed.

But still, my confusion is predicated on the assumption that people want it quickly, and you blatantly don't seem to be be one of those people or you'd have it already anyway -- so the fact that you didn't bother is perhaps less surprising than it might be.

Again the day of a Fedora release comes, and a whole bunch of people suddenly start downloading the whole thing from scratch, bringing the mirrors to their knees.

I've never really understood this behaviour. Do people not understand that the rawhide development tree is fairly much identical to what actually gets released? Do they think we spend months stabilising rawhide and then just for a laugh we decide to release something entirely different?

All but one of the final FC5 packages have been available from rawhide for about a week now. The only difference is the 'fedora-release' package which contains /etc/issue and the yum configuration. That's one RPM out of 2369 (for the PPC version). Just over 1M, out of 3.4G.

To update an install tree (or indeed an installed machine) from rawhide to FC5 with rsync would take moments... yet there are people out there who have only just started to download all the other bits too.

It's weird -- I can understand waiting until a few days after the release if you're not that desperate to have it, but to be waiting for the moment the mirrors make it available and only then start to download it all is just odd.

2 Mar 2006 (updated 2 Mar 2006 at 13:00 UTC) »

I took a quick look at the bandwidth and latency graph for my home DSL line -- Evolution's misbehaviour is fairly obvious on it.

I started the 'new and improved' Evolution at about 7pm on the first day, and it spent about two hours downloading headers before it fell into a pattern of thrashing the link once every few minutes, downloading the flags of every mail in the folders it was checking.

Note that this was after I'd applied some of my own fixes to it, to make it check mail only in active folders rather than in all folders. Without that, the problem would be about an order of magnitude worse, because it would do have been downloading flags for about 3 years worth of historical archives each time, rather than only for mails in the currently active folders. Out of the box, Evolution would probably have given me a solid mass of green in the graph.

On the afternoon of second day, you can see that I stopped it -- that was when I was playing with the patch which I then submitted to Red Hat bug #183219. After an hour or two I gave up on pursuing that approach myself and restarted the original -- you can quite clearly see when that happened too. I looked at the bandwidth graph and then killed Evolution at around 12:45 on the third day, and then the line usage went back to normal.

I'm now looking at kmail; once I add the facility for accessing an IMAP server over SSH, it should do quite well as a replacement. In fact, if I'm going to start using KDE programs I might as well just switch over to KDE wholesale -- it'll save me from the constant temptation to kill someone when I have to deal with the insane GNOME file dialog box.

Looks like the Evolution hackers have been smoking some really good crack recently.

In its folder tree, Evolution shows folders with unread messages in bold, with the number of unread messages in brackets after the folder name. A relatively sane and useful thing to do.

Once upon a time, Evolution would check how many new messages there are in each folder by issuing a STATUS command -- a relatively quick transaction which looks something like this:

--> 000 STATUS lists.random.goatsex (UNSEEN MESSAGES)
<-- * STATUS "lists.random.goatsex" (MESSAGES 221 UNSEEN 221)
<-- 000 OK Status completed.

Now, it wasn't particularly clever about this -- it would do this for every folder on the server, even folders which were inactive and which never actually received new mail. And it would do this to the detriment of all else -- ignoring user commands to actually fetch mail to be displayed, change folders, etc. until it had finished doing this check on every mail folder. So if you're on a high-latency link it would appear to go off into the weeds for minutes at a time, failing to display mails which you click on because it's busy doing other things you don't immediately care about.

But now it's worse. Instead of the STATUS command, Evolution now issues a SELECT to actually open each mail folder, then proceeds to fetch the flags for each and every message in the folder and also even the headers of each and every message in the folder, if it doesn't already have those cached. Then it iterates over its list of mails, counting the ones which have the 'unseen' flag set. It comes up with a number, which it uses to display the unseen-count in the folder tree.

I have to offer my congratulations to whoever made this change -- you've turned a three-line IMAP transaction into a multi-megabyte download taking many minutes, for every mail folder on the system. When I started Evolution for the first time after 'upgrading' it, I had to leave it overnight to let it finish starting up and doing this check.

Now put down the crackpipe and back away from the editor...

Throw away your television
Time to make this clean decision
Master waits for its collision now
It's a repeat of a story told
It's a repeat and it's getting old

124 older entries...

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!