Peter Flynn

I have Ubuntu on almost all machines except a couple of big RHEL servers and the domestic Macs. The oldest desktop is a Dell something-or-other with a 1024×768 screen but it's been happily running 8.04 LTS (Hardy Heron) — so why would I want to upgrade?

Partly I was getting tired of needing to use a new facility or test a new package, only to be told that it required more recent libraries than I could install without major internal surgery. Plus the upgrade tool was rather explicit that 8.10 and later were unable to support my nVidia Geforce4 graphics card because the existing driver had been obsoleted by changes in the Xorg code; but that my installer would automatically switch to the older (FLOSSy) nv driver which would at least get me running. Everything was backed up, so I took the plunge.

They either lied, or something went wrong that they were unaware of. The upgrade went right ahead and installed the new driver which it knew was incompatible, so the newly-installed system simply hung at the login screen. Several hours of poking at the drivers convinced me that they just hadn't bothered to check. Several more hours of emailing the CLUG convinced me that it wasn't worth repairing the damage. I was resigned to reverting to 8.04 until someone mentioned 9.10 — I'd been looking at it, but had assumed that if 8.10 had dropped support for my graphics card, there wasn't an icicle's chance in hell that 9.10 would have put it back.

Wrong. I wiped Intrepid Ibex and installed Karmic Koala and it came up first time, full graphics mode, and even Compiz is working (so eat your heart out, Snow Leopard). Lots of good stuff, a few stupidities (not Ubuntu's fault): no KPDF, no KDVI (KDE has integrated the first into KDEgraphics and apparently dropped the second on the floor; but they have removed the ‘Current’ [page] option from the Print menu, which is plain daft); and some prat has removed the list of current buffers from the Emacs ‘Buffers’ menu — what am I supposed to do, guys? remember all my files in my head? C'mon, put it back.

But there goes the doorbell; kids wanting another blast of the pressure hose, I guess…Happy New Year to you all.

Saturday 2009-10-31 20-02-00

We've been typesetting a book for a colleague of mine, a Festschrift for a colleague of his. As usual with these works, each chapter was contributed by a different author, and my colleague had the task of putting them together.

We consulted about it beforehand. He uses OpenOffice (perhaps NeoOffice, I can't remember) on a Mac, but he had taken in all the chapters and edited them and arranged them, and was pretty much ready to go, when the publishers reminded him that their policy was for endnotes, not footnotes. Not a problem: only Chapter One used them, and we were using XSLT to create LATEX code for the formatting, so no changes were needed except to switch in the endnotes package.

However, which checking Chapter One, my colleague observed that all bar one of the footnotes-now-endnotes just said ‘My italics’. The author was quoting extensively from other sources, and highlighting relevant words and phrases. My colleague agreed with his publisher that such repetitive notes really weren't needed, so he left the only ‘real’ one and deleted all the rest. But as he deleted the final endnote of Chapter One, all the rest of the book (twelve chapters) vanished.

Ever a man of cool mind, he hit Undo and brought them back. No harm done. But that final endnote of Chapter One was back as well. Assuming that he had slipped a finger, he again deleted it…and again all the remaining chapters disappeared. At this point he brought them back with Undo, and saved the file for me to look at.

Well, well, well. The first time I had had to dig into ODF XML in earnest. It's nowhere near as bad as OOXML (so no surprises there, then), but footnotes and endnotes are inside more multiple containers than a Russian doll, and that was where the problem lay.

ODF allows multiple paragraphs in a single footnote/endnote — perfectly reasonably; all general-purpose markup systems do that, from DocBook to LATEX. But in ODF, as in OOXML, everything is a paragraph. Neither system has any concept of depth, regardless of the repeated and redundant nested sections in Word documents and the Byzantine sewer of ODF's footnotes.¹ That means ODF (and perhaps OOXML; I haven't tried it) allows anything at all in a single footnote, even twelve whole chapters because everything is a paragraph, and the only evidence of the text being in chapters is the style name, which is not checked because it's not a controlled vocabulary.

And that's what my colleague had unwittingly done. Having edited the first chapter, his cursor had been at the end of the last footnote, and had remained there while he pasted in twelve more chapters, which promptly and silently went into the footnote container and stayed there.

The reason is clear: OpenOffice believes that the cursor has no business being in element content at any point after the last text node, so there is actually no way (that I have found) to insert a new paragraph sibling to your current ancestor paragraph when your cursor is at the end of a footnote.

This slavish adherence to the WYSIWYG model is fine for shopping lists and business letters, which are short, transient, and not subject to any reprocessing. It's not suited to a serious editor, and if OpenOffice wants to be taken seriously by the publishing community it is going to have to change this, and preferably introduce a style margin like Word (or some other wide-angle style view) so that editors can see what they are doing. The last time I asked them, they said you could see the style by hovering the mouse over a paragraph (clearly not someone who had ever had to edit large documents), which is not meaningful when you need to see perhaps a dozen paragraph-level styled elements on a single screen.

It also makes a fine example of another technique I am researching for easing the burden on writers, and that's the distinction between ‘Insert’ and ‘New’. The first is for database programmers, markup experts, and ontology gureaux. It's meaningless for writers because with only a WYSIWYG view, there is nothing to insert into (and the interface won't allow the cursor to be in element content). ‘New’ is what writers use (new chapter, new paragraph, just like you were dictating), and when you want a new one, the editor should check its current location, scoot forward to the next place at the current level where one is permitted, and add it there, creating additional markup if needed to retain validity. If there is no place at the current level, go up a level and check there; lather, rinse, repeat.

The facility for an editor to click on ‘New Chapter’ and have the editor Do The Right Thing, instead of the wrong thing, would be a major step forward.

1. One exception: ODF wraps lists in proper list markup. Well done, ODF.

Saturday 2009-03-07 21:41:00

Every so often, ISPs have a major purge of hardware and software, and ‘upgrade’ their clients to the new service. You don't get a choice, and you're lucky if you get any warning, because alerting the clients that there is about to be a chnage might lead to some disaffection.

I've been with this bunch for over a decade, and they've given excellent service, with a responsive helpdesk and virtually no outage or downtime. The service was simple, and as most of my sites are generated rather than hand-coded, this suited me just fine.

Two weeks ago I got email that warned me about some small changes in their mail systems during an upcoming upgrade. No problem: I downloaded my mail folders and switched from IMAP to POP for a week (easy when your IMAP, POP, and webmail service all use standard mailbox files: goddess help those whose mail is buried in proprietary databases). After a week I uploaded the mailboxes and reverted to IMAP.

Suddenly this week everything stopped working: mail wouldn't log in, not by any protocol, and all my sites were inaccessible — but oddly, I could still log into their new site-owner page, and FTP still worked. I logged a support ticket and downloaded a few key files. No word of any changes, though.

Then the FTP access changed: I could still log in, but the entire directory structure was different, although all my files were still there, but in a different subdirectory. No response to the ticket. Phone calls met the usual disembodied voice saying that all their agents were busy helping other clients, and refusing to put me on hold, but offering to take voicemail. No response.

Fortunately, their web site gave local phone numbers (they're in California: I'm in Ireland, so an American 1-800 number is no use to me). A phone call early in the morning Pacific time got a harassed support person who explained they'd been having ‘a little difficulty’ in the move to a new hardware and software system, and claiming that they didn't have my alternate off-system email address (which they did have a week before because they emailed me on it). All was eventually resolved and the email and subdomains re-established (except I lost a dozen or so planned-but-nused subdomain names). Generally speaking, pretty good service, although it would have been better if they'd told me beforehand.

It was worrying that the move clearly squashed all record of my subdomains, all my contact details, and reset all my email passwords. But what was reassuring is that they had the good sense to put a non-1-800 number on their web site, and they didn't change my owner password. Everything is backed-up off-site anyway, so I wasn't worried about data loss, and because everything uses open file formats, a new environment is no big deal. To those of you who went looking for the XML FAQ, the Acronym Server, or the online book Formatting Information, thank you for your patience.

Friday 2009-03-06 21:26:00

One of the recurrent problems in living on an island off the coast of a bigger island off the coast of a continent is that you have to use boats and planes to get anywhere other than your own doorstep. Planes are not a problem: our local airlines do a reasonable job, and RyanAir terrifies the life out of bigger airlines worldwide.

But if you want to bring the car, you either have to drive for hours to get to a port for a ferry to Wales, or wait a week to get the ferry to France. Either way you pay through the nose for the privilege, because the shower who ended up running what was left of the direct ferry to Swansea lost their tub a couple of years ago and never found a new one. What a pity the proposal to the (then) EU by our former Foreign Minister Peter Barry for a bridge or tunnel to the UK never took off!

However, at last there is new hope: a local campaign to bring back the direct ferry service is making excellent headway. If you're interested in coming via the UK to visit, or the other way round, sign their petition and fill in their questionnaire, and maybe we'll soon be back in touch with the next chunk of land.

Monday 2009-01-26 20-35-04

The recent XML meeting was in DC in December, and I was pleased to see more evidence of the developments I mentioned above, although the whole meeting has shrunk (as predicted) as XML becomes embedded into the wainscoting of IT development and ceases to be new. The next breakpoint is the Balisage meeting in Montréal in August (the week after Worldcon), and I hope to have available some results and conclusions from the work I have been doing on the usability of editing software for structured text.

The TEX Users Group meeting last July was well-attended, and went without any major hitches except one: the odd inability of the host institution to provide visitor wireless access. Curiously, this was not only of no interest to UCC, but they seemed to be unable to understand its importance to visitors. It does mean that organisations looking for a host site would do well to check wifi arrangements and do a physical test that it exists and works before committing to the site. Fortunately this year's meeting in Notre Dame is better-prepared in this respect.

There's a load of new stuff to go through in the next few weeks: an updated Maemo, some neat Yuletide presents, lots of books, and plenty of file-format weirdness.

Monday 2009-01-05 18-22-46