Gallimaufry of Whits
Being for the Month of 2007-06
These are quick notes taken by Sean B. Palmer on the Semantic Web, Python
and Javascript programming, history and antiquarianism, linguistics and
conlanging, typography, and other related matters. To receive these bits of
dreck regularly, subscribe to the feed. To browse other
months, check the contents.
The Pamphlet Saga so far: Pamphleteering and
More Pamphleteering.
Today has been a day of Countryfile, chatting with Terje about Blue Moon, and
the like. These sorts of things often make me think about the Pamphleteering
idea. I really need to boil the concept down into its most useful constituents
and give it at least a preliminary name so that I can start to hype the crap
out of it. I'm really stoked about this idea, even half a month since
conceiving it.
I've already got 5000+ words written about the idea, and that's just a brief
overview of the whole thing. William said that after scribbling down notes to
myself I should think about writing them for one or two other people, and that
sounds like a pretty good stepping stone.
I set a Swhack
Challenge today: find the original source of Lagrange's famous comment about
Newton. Not long after, I found the answer
myself: it's from Jean-Baptiste Delambre's eulogy of Lagrange printed in 1816.
Translated into English, thanks to Robin Berjon and Daniel Biddle, the phrase
reads something like,
"also Mr Lagrange, who often quoted him as the greatest genius ever to have
existed, added immediately: and the most fortunate; one only finds once a
system of the world to establish".
Charles Goodier, "mediovia", mentioned today
in passing that he's reading Whits, which was a nice surprise. Hello Charles! I
met him years ago at the WTF1
conference in London, and he gave me a disk of the Wikipedia dump that he'd
kindly prepared for me. I still have it... you don't forget an introduction
like that. "Oh hello. Here's that huge amount of information you asked
for."
Charles is working on the finance side of things with ESP, which I've
mentioned once or twice. The current state of affairs with that is that Tav,
who started the 24 Weeks project, has bumbled off to Spain until the 18th, and
so people have been left to regroup a bit in his wake. I've gone back to
playing neutral observer; there isn't a huge amount to observe, but it's an
interesting time of mutiny on the high leas.
Of course I'm distracted anyway with the camera that I mentioned
back in February. It's always fun when there's a change of routine since you
can dispel old habits. BreakBrokenTraditions and all that. The manual mode is a
lot easier to use on this camera, so I've been playing about with all the
settings and I'm taking some pretty stonkingly good photos. I've got macro down
pat, and outdoor shots are getting pretty good. Indoors is a bit hit and miss.
There's a slight problem with the centre weighted zoom in that it's not all
that... centre weighted, so I might switch to point zoom and see if that's any
better. Apart from that, it's proving great fun. I love photography.
To readers of Lo and Behold!, if there be any: I haven't forgotten.
I've got some cracking ideas, but they're not coming together at the
moment, so I'll probably have to switch to Plans B meanwhile. I'm still
planning to get the first part of the serialisation published this month, after
the current busy spell, which is reducing my outputs on various fronts such as
Lo! and Whits, but probably increased overall from the previous few months,
which themselves were quite productive. Anyway, I haven't forgotten about
Lo and Behold!
Schuyler read my
thing about Stable P2P.
The idea of it was that using the web for storage is easier and better than
using a P2P system where all of the files are stored persistently. Schuyler
pointed out that for one use case, that of a large non-profit organisation,
this isn't true; and that bandwidth is another interesting consideration.
We also came up with the parallel of a SETI@home type project only
harvesting disc space rather than CPU cycles. Might be a good way for the large
non-profit organisations to get their needed disc space in a distributed but
non-peer environment.
In talking about capacitors, squirrels, and eventually horses with James
Arthur today, I remarked elsewhere that it was a right "carnival of thought".
Searching about for that almost novel phrase led me to this wonderful sentence
from The New Monthly Magazine edited by William Harrison Ainsworth
(1855), Volume 104:
"M. Chasles tries to awaken a becoming interest in his character, and
curiosity as to his style—that chaos of parentheses, ellipses, and latent
meanings, that carnival of thought and language, that labyrinth without an
Ariadne's thread, that mingle-mangle of impracticable events, impossible
geography, unaccountable characters, of quotations, interjections,
exclamations, puns, epigrams, impertinent episodes, abrupt discords,
measureless digressions, merciless divarications."
Chasles was talking about Jean Paul, "a German writer,
best known for his humorous novels and stories". Might be a fellow worth
checking out!
I've put up some Design
Stuff, including some CSS and Javascript tricks, web tests, and some
miscellaneous interesting stuff like a Java (the country) Train Timetable for
spies that Tufte mentioned somewhere.
On a MacBook keyboard, en-dash, –, is Alt+-, whereas em-dash, —, is
Shift+Alt+-. This is logical, but not huffmanised! Huffmanisation should
probably win in this case. Of course, given that the MacBook keyboard has a
whole key devoted to the section symbol but # is hidden away as the very
annoying Alt+3, there are worse problems that need fixing...
Gran
Paradiso, Firefox 3.0a5 was released today, so I tried it out. It was
sorely unimpressive, slow and buggy without any noticable feature changes over
1.5. This got me on another of my browser migration wonderances, and I tried out Safari,
Camino, and eventually Opera with various problems and annoyances along the
way.
I didn't settle on migrating to Opera, but I was intrigued by its Speed Dial
feature. There's an extension for Firefox that implements Speed Dial, but I
thought that I could go one better. My current start page is a list of some 30
links or so, to general services and the like. They're all text links, and I
only use about ten of them with any regularity, so I thought it'd be neat to
get logos of those sites and put them on a Speed Dial like page.
The results are
quite cool; the principle of having a larger target for more frequently used
items is followed in comparison to the old small text links, and it looks
great. I had to hunt about a bit for some of the logos: the OED one is a third
party logo; the Froogle UK one is the old logo that someone's archived
somewhere (I had to resize it, and others). The Gandi one was a transparency,
so I had to set the background colour. Apart from that, it was quite
trivial.
This got me thinking again about having a start page for creativity, though.
This new start page, and indeed my current one, are motivated towards letting
me find sites quickly when I have a task in mind already. I figure I want to
buy a book, so I go to Amazon. I want to find out where something is, so I go
to Get-a-map. The tasks are decided before I go to the page. But what about
when I'm trying to generate ideas and be creative? How can the web help me
then?
The only page I've thought of so far that could go on a Creativity Start
Page is Wikipedia Random Page; but I've tried using that before to spark some
creative idea, and to be honest it's not all that good an approach. Does
anybody else have any ideas? Perhaps I should link to lots of things that I'm
interested in... but random queries of interesting areas (Notes & Queries?)
seem to be the way to go. That might require some scripting magics.
"Prove it in Python!" - The new #esp motto.
When you move a page from somewhere.html to somewhere/index.html the server
automatically puts in a redirect for you usually (it adds trailing slashes to
things for you), but if you do the reverse it won't strip them off. Here's
something that I came up with a while ago that will achieve that in Apache:
RewriteEngine on
RewriteCond %{REQUEST_FILENAME} !-d
RewriteCond %{DOCUMENT_ROOT}/$1.html -f [OR]
RewriteCond %{DOCUMENT_ROOT}/$1.txt -f
RewriteRule ^(.+)\.(html|txt)/$ /$1 [R]
I implemented it on infomesh.net when I found that I had quite a few things
which only consisted of a single index.html file in a directory.
Today I took photos of a jellyfish and a swan. They came out
pretty well; those were just two from a set of 150 or so. I've used Flickr
partly to just try it out, partly to give back to it since I've used it fairly
often, and partly because the originals are quite large. On that last point,
however, it turns out that Flickr requires you to upgrade your account
(£12.65/annum!) to let people download the larger photos, which I'm not going
to fork over given that I've only uploaded a couple of photos.
Christopher Schmidt, who just got married (congrats!), says that they'll
retroactively enable the hi-res downloads if you upgrade though, so I can use
it intermittently before deciding, which is nice. Björn suggests using www-archive which
would be funny for one or two photos, but is probably not a good long term
strategy!
I must needs petition my library, for it has erred. I went to browse the
Notes and Queries collection, as I often do, only to find that it had been
moved to a new section. Okay, I thought... But when I arrived at the new
section I found that the pre-1950s editions weren't there! I checked in the
catalogue and they had been moved to holding storage.
So I went to the enquiries desk to ask what had happened, and they said that
they'd been reorganising to make some more space, and had therefore moved some
stuff like the old Notes and Queries collections to storage. I pointed out that
I use the collection regularly and they gave me a suggestions form to fill out,
by which to make my petition. The Notes and Queries collection was really one
of the best things about the library, and it's an awesome library anyway. Let's
hope they listen to common sense eventually!
It is possible, of course, to request things from storage, but I think you
have to ask in advance, so it rather precludes the whole browsing workflow
pattern. And anyway, it's an unnecessary tax on both them and me, so hopefully
the quick suggestion will do the trick.
Reading E.K. Chambers's Facts and Problems II this morning, I came
across one of those many delightful little quotes, this one of Dryden
recounting Jonson. Oddly, though, it's only reproduced on a single webpage before this
one:
"In John Dryden's Essay on Dramatique Poetry of the Last Age
appears the following passage: 'In reading some bombast speeches of Macbeth,
which are not to be understood, he [Ben Jonson] used to say it was
horrour.'"
I wonder when Chambers goes into the Public Domain? Wikipedia says that
anything "before January 1, 1923" (or 95 years ago, whichever), or where the
"last surviving author died at least 70 years before January 1 of the current
year" is generally public domain. Facts and Problems came out in 1930,
so it's not been those 95 years yet. E.K. Chambers died in 1954, so it's only
been 53 years. That means we have another seventeen years to go... 1954 + 70 =
2024. Disappointing!
His Elizabethan Stage came out in 1923, though, so there might be
some stuff surrounding that which is publishable.
"This leads to the obvious desire to define a constraint based grammar for
specifying home design" - Anselm, via Chris.
Björn was telling us about his awesome grammar engine, Alexander (a kind of
RNG-for-strings), and one of the things that he noted is that he can provide
complements of regexps. So for example, a regexp of all the strings which
aren't URIs. I guess that sent me over the edge: I know that Anselm probably
just meant interior design, and he was probably just joking, but there's
something about using regexp to generate homes which seems like a good
idea.
So for example, you might specify that the lounge and the kitchen must be
adjacent, as a rule. But what constitutes a lounge, and what a kitchen? There
you might have further rules: a lounge must be at least 21 feet long, and have
at least one window which is over 5ft wide. And so on, since obviously there'd
be all sort of rules for adjacency of windows, their height from the floor, and
so on.
The aim is to get there to be enough constraints so that you can use a
genetic algorithm or something to generate a load of house designs that conform
to the constraint, and then to use that to gradually hone in on the best
possible house that matches your house grammar. This depends on the
requirements, i.e. the grammar productions, being correct, but for conservative
noddy boxes that's probably not too much of an issue—there are probably quite
hard and fast rules about things like how low to the ground you want to put a
window. If not, perhaps people might end up sharing them.
One way that this might be worked is to have a mini-language that generates
a 3D boxy object, with adjacent walls being a bit like adjacent nodes in a
graph or whatever. Windows and doorways could be like attributes... Then you'd
just end up writing a schema for that language. It might end up being more like
schematron than RNG, because houses don't really have logical roots; they're
really going to be more like graph structures than trees. Seems like an
interesting idea, though. I'll bet somebody's already done something like this;
too good an opportunity to have been missed until now. Perhaps with not quite
this amount of compsci flavouring though.
Some more photos for your delectation, this time of a Thistle, Bugs and Butercups, and
a Giant Daisy. All
three of these are essentially macro shots, whereas the last two I posted were
merely closeups.
I think I've got a little theme going here: common natural phenomena well
taken. I mean, a swan, jellyfish, thistle, buttercup, and daisy aren't the sort
of thing that you expect people to take photos of these days... it's kinda back
to basics.
Behind the scenes, these three were selected from something like 390 photos
that I took that day, the majority of which were landscape photos. Generally
I'm taking photos from the top 10% but not the top 5% in quality. Going through
hundreds of photos selecting ones to post is actually quite a longwinded
process; I'm definitely thinking about getting Aperture.
The Whits Word of the Day: "ombotrophic" (fed by
clouds).
I have lots of little ideas for fiction. I was wondering if I should collect
them all together somewhere and have that be the fiction, with a very murky
line between what is canon and what isn't. Structure is the only canon! The
problem with fiction I find is coming up with characterisations that are both
compelling and useful. Fiction is usually quite an umbrella discipline,
allowing you to use the act of sub-creation in order to work on various things
that you wouldn't normally think about. I really like umbrella disciplines...
though I haven't thought of a great many really handy ones.
The Ghyll wiki was
interesting for being creative and yet rigorous: the only way that you could
play things out was by describing them fairly dispassionately. Actually there's
a lot of scope for bending the rules in fictional encyclopaedia entries.
At least I wrote up the metaschema here!
Today, a post on Lo and Behold!: On Midsummer's Night. It's kinda
ironic that I should start writing the Summer 2007 edition right in the middle
of summer, but of course that depends somewhat on how you reckon the seasons,
the subject of the second essay there. The cool thing about them so far is that
they're actually quite practical and concentrated, for all their density.
Typesetting the first 7000 words that comprise the Spring 2007 paper edition is
going to be quite a challenge, though.
Dave Pawson contacted me with a
question about modelling confidence in RDF. I'd done some work on this in EARL way back
in the day, but the confidence stuff was recently removed from the EARL schema
I see.
An insight that Dave came up with is that confidence isn't a property of
information, it's the property of the source of the information. So for
example, say you're told that Paul Brown was born in Rochester. The source of
your information is Paul's friend, Michelle. You'd say { :Paul :birthplace
:Rochester } :says :Michelle, and then you can say :Michelle :trust "3".
The problem with this is that you might trust Michelle middlingly on this
information, but more on another piece of information. So you'd end up with
:Michelle :trust "1", "2", "3" etc., which doesn't make any sense. Clearly,
confidence is at least a property of the utterance and the utterer. So you
might have { { :Triples :terms "3" } :says :Sean } :confidence "3" . ("I, the
writer of this document, kinda believe Sean when he says that triples have
three terms") and { { :Triples :terms "4" } :says :ManInTheMoon } :confidence
"2" . ("I, the writer of this document, barely believe the Man i' th' Moon when
he says that triples have four terms").
Dave appears to believe that it can work on a layered scale. So you might
have a default trust level for a person, but then you can add information about
a particular utterance by that person later. So say you have { :Triples :terms
"3" } :says :Sean . and :Sean :defaultTrustLevel "2" . in a document. You might
derive from that that { { :Triples :terms "3" } :says :Sean } :confidence "2" .
unless you have something specifically saying { { :Triples :terms "3" } :says
:Sean } :confidence "3" .
Layering the confidence seems like it might work to me, but trust is a
rather dificult area when it comes to modelling. When you increase it to the
size of a trust network, you're in for issues. The rather interesting discussion
surrounding this happened on #swig with me, Dave, Dan Connolly, Jan Grant, and
Dan Brickley involved.
Firefox annoyed me
to the point of actually switching today, so I switched to Camino. I'll
probably switch back because I'm so used to the Firefox workflow, plus Camino
has some, shall we say, idiosyncracies of its own (no undo-close-tab, and
non-displayed-tabs when the tab bar overflows).
What does a browser actually even do? Not very much, if you consider it as
just a wrapper for the rendering engine. Rendering engines seem pretty good
these days; not perfect, given that they almost all still fail the CSS 3 selectors test,
but most of the practical problems that I face don't seem to stem from the
rendering engines.
A browser itself seems to provide the following wrapper facilities around
the rendering engine: navigation, tabs and windows, bookmarking, history,
search widgets, control over Javascript fonts and other features of the
rendering engine, form filling in and password remembering, and... that's about
it. You get some more wacky stuff like RSS readers and so forth, but I tend to
think that applications should stick to the unix mentality on this sort of
thing. Do one thing, but do it really well.
Well, okay, not quite the unix mentality since there are several really
essential features that a browser provides, but it occurs to me that in actual
fact many of them are facets of a single thing: navigation. Bookmarking and
history, I mean, kinda get subsumed into navigation when you think about it
from a higher perspective.
Consider what a browser effectively does as its core task. It shows you
pages on the web, from a user specified entry point: clicking on a bookmark,
clicking on a history item, entering a URI in the address bar, clicking a link
in a webpage, being passed a URI from an external application. When this page
renders, the page is saved in the browser's history. The act of navigating to a
page in one of these ways and rendering it is an act which is recorded in the
history.
But not only is it saved and recorded in the history, if you specify it to
do so it also saves it as a tab. That is, you can render documents at parallel
times and have easy access to them. I'm pointing out the obvious here to show
something that is perhaps not so obvious: tabs are really a kind of strange
history. They're a history where basically all of the members have been cached
for easy access, and you have a nice nagivation bar exposing this particular
cached history right in the browser window. They're stacked in the order in
which you opened them, but in Firefox 1.5 onwards you can reorder them. This
tab paradigm doesn't actually have a persistent history itself, but we do find
such a thing useful, so we bolt it on in the form of session savers. I also
find myself bookmarking all of my tabs. Tabs are really a slightly scummy (not
persistent) and yet also advanced (more easily accessible and cached) form of
the browser's history. And yet because it's a separate system from the actual
history, it's considered orthogonal. You open a tab and an entry gets put in
the history. They work side by side.
Bookmarks are more like the browser history than tabs. Bookmarks are kinda
selected members of the history, with a user based structuring; that is to say,
basically a hierarchy of folders. It's a tree structure, though some browsers
are now allowing keywords. You can have multiple entries in bookmarks just like
you can have multiple pages in your history (if you've visited a page two or
more times). But the thing to remember is that bookmarks are user controlled in
a more flexible way to the history: with bookmarks you can go into the
structure and change things. And nothing gets added to the bookmarks as a
consequence of another action, unlike history, so it's less automatic
in that sense.
But what about navigation itself? When I click a link in a webpage, what
I've done is to traverse a link on the internet. But in my history that just
shows up as another entry. There's no link between the two items. There's also
nothing that tells me the difference between typing a URI in the address bar
and having been sent a link via an external application such as X-Chat Aqua. In
other words, the browser history is descriptive—it's a record of the pages
which are being visited in the browser—but it's selectively descriptive. You
might say that it's imperfectly descriptive. Another thing that the history
doesn't record is when you use the back and forward buttons in your current
tab. It also doesn't record when you switch between tabs. What about if you
could have a split screen of tabs? Then the topology to describe would get even
more complex.
The history is a descriptive structure, but not descriptive enough. The
bookmarks are a prescriptive structure, but perhaps not flexible enough, and
not all that tightly coupled to the history. Say I have a bookmark and I want
to find out all instances of that bookmark in my history. Can I do that easily
in a browser at the moment? As for tabs, they're a kind of strange stunted
prescriptive-oriented history. I can go in and change the structure and reorder
the tabs, just like with my bookmarks. The main difference is that tabs are
more easily accessible (or, actually, it depends where you place your bookmarks
toolbar and if you even have one), and that most of all they're caching the
recently gotten content of the page and also internal state such as text in
forms and any changes to the DOM done by Javascript or whatever. There could be
other forms of caching... you might cache just the original text of the page
but not its rendering. A history is kinda like a cache of just the URIs and the
titles, coupled with the time (but not the mode) of access.
It seems quite clear that these three things, history and bookmarks and
tabs, can be improved and unified to some extent, and that other similar
paradigms may be possible.
As for the other features, navigation and saving form information and the
like, these can probably be improved upon significantly too. Camino doesn't let
you save multiple input information for a form, which is just terrible. There's
some application for OS X that actually lets you save the most common bits of
inputted data, which I think is a pretty good idea. Things like that should be
available everywhere.
And this is just the structural side of things. When you throw the UI into
the equation, you've got a whole new set of ideas that can be explored. The
placidity of browser designers these days is horrific. Browsers are engineered
terribly, with a bad selection of the best default features, and an
overproliferation of options that people don't need to commonly access. I hope
that there's another Phoenix-like revolution at some point within the next year
or two, but it doesn't seem likely. Even though a browser is mere cruft around
a renderer!
Arnia and I discussed the semantics
and behaviour of navigation in browsers somewhat more, with some interesting
ideas.
I'm considering switching Whits to using a more Lo!-etc.-like structure, so
that each post has a shortname. Thus, instead of /whits/2007/06#N231507 for
this post it'd be merely /whits/shortnames. One drawback to this is that then
I'd have to consider that each post must make sense out of context. It does,
however, make them easier to link to. I'd aggregate all the month's posts into
YYYY/MM pages still, too, so it'd be backwards compatible, not that it needs to
be.
On my happening to mention
that "I just fail to get excited at pretty much everything Semantic Web these
days", I had a good chat with TimBL about some qualms I have with the Semantic
Web, and some interesting hints of answers as well as (and in the form of) Tabulator development
breadcrumbs.
One of the biggest things is that Jim Hollenbach has created a Tabulator
Firefox extension. Though I
link to it, there's not much point in trying to install it yet: I've tried with
the help of both TimBL and Joe but there's stuff that needs fixing for it to
work in Firefox 1.5 and 3.0a5, and even on 2.0 you need to be able to get the
latest development tree of Tabulator and graft it in at the right place. All
the same, the extension idea is one I had last year or so and this progress is
certainly encouraging.
Another idea that I had for the interface was that rather than simply
browsing around by clicking, some kind of text input would be nice too. Some
kind of RDF path derivative, in fact, which allowed you to do commands as well.
So for example, one use case I have in mind is updating my FOAF file. I might
load it, and then want to traverse to some person and get to their data. I
might want to then load a friend from their FOAF file who is now also my friend
thanks to their introduction, and load some of their details (perhaps just name
and mboxsum) into my own FOAF file. There are probably parts of that which are
quicker with typing and parts which are quicker with drag 'n' drop and the
like. The thing about Tabulator, as TimBL observed, is that there's so much
potential with it for things like this.
Today Danny Bruder put together the latest radio show that I scripted for
him, and the
results (33 MB) are pretty much just as you might expect a radio show
written by me and produced by Bruder to be. What are you waiting for?
There's also the script
itself for you to follow along with, if you like.
Louis Jordan recorded a song called Ain't That Just Like a Woman in
1946. The riff at the beginning was lifted by Chuck Berry for Johnny B. Goode a
decade or so later. Surprisingly, or perhaps not, there's a video on YouTube of Jordan
performing Woman on the stage... it looks like he's miming it, but he
deviates from the recorded version in places, including the bit that goes: "You
read your history books, you've learned that Nero fiddled while Rome was burned
/ Ain't that just like a woman? Ain't that just like a woman? Ain't that just
like a woman, they'll do it every time!" Most of the verses are actually about
women, so I'm not sure that the Nero bit is about.
The film that the video is from is public domain, so I might try to excerpt
an MP3 of the performance from it and stick it online. It might be nice to make
a little collection of public domain music that I like if I can find a
reasonable amount. Probably not, but then this was unexpected (if it turns out
not to be false).
A cartoon about removing
the apostrophe from English got me thinking about reverting to Early Modern
English spellings again today. I really like Early Modern English spellings and
orthography. It's not too difficult to learn, so conversely it shouldn't be too
difficult to write a guide. I'm also thinking about creating a script for
automatic conversion of at least the most general things. Going back a bit too
far, though it is in a piece about Early Modern English, I found
a nice passage about pre-GVS English: "Chaucer lived in what would have sounded
like a hoos, with his weef, and hay would romance
heer with a bottle of weena, drunk by the light of the
moan."
Things like Ben Jonson's quote on Macbeth just
wouldn't be the same if related without Early Modern English spelling
conventions.
Sean B. Palmer, inamidst.com
