Re: [Server-devel] Wikipedia on XS

2010-07-27 Thread Chris Ball
Hi,

I think Martin's answers are all correct.  Just to be clear:

Strange -- it does have images for me. Just tested -- from the
homepage, I click on the Sociology link and on that page I get
the picture of Auguste Compte. Both Wikipedia English and Spanish
carry it.
   
Hmm. Or do they? The images are actually being referenced from
the internet. This is a regression from earlier releases of WA.

Yep.  We didn't make an image selection for the English snapshot, we
were already very article-starved trying to fit in 100MB, because the
English WP articles are on average longer than the Spanish ones.  (And
of course there are many times more of them.)

So, no images in the English version.  The Spanish version contains
some (3000) images, and will call out to the net if available for
the rest.

When I compare WA English with WA Spanish I see the Spanish one
contains a good number of images, where the English one doesn't.

Right.

* The Wikipedia Activity links to articles not included in the
activity.  The schools Wikipedia does not include any links to
articles not included in the Schools Wikipedia

Correct -- they are usually marked with a different colour
however...

Right.  The links that aren't present in the archive are still shown,
so that you could click on them if you happen to be connected, and are
a different colour, so that you know which ones to avoid if you aren't
connected.

* The Wikipedia Activity is very slow on an XO-1. The Schools
Wikipedia is very fast with good wifi to an idle schoolserver

Wikipedia Activity is highly compressed, so it's unzipping things
behind your back.

Right.  (Actually, the slow parts are mainly converting from wiki markup
into HTML, and performing template expansion, which involves recursive
references to other articles.  A single article might end up pulling
fifty articles from the archive, forty-nine of them being templates.)

The Schools Wikipedia is already in HTML, so it doesn't have the
decompression stage, the template expansion stage, or the converting
to HTML stage.  It trades-off for using much much more disk space.

* The Wikipedia Activity has a lot of articles that might be
considered inappropriate for some or all age groups by some
cultures. I haven't found anything in the Schools Wikipedia that
I think might be considered inappropriate, but it's search
features are poor and I haven't done a particularly thorough
search.

There's been some editing in WA but I do believe you might find
tricky topics

Yes, we only removed articles on (specific) pornography and some sex
acts.  A deployment would have to decide to reduce the article set
farther themselves; we weren't comfortable limiting other types of
knowledge for everyone.  It's all available on the main Wikipedia
site in any case.

Given that your planned XS is fairly powerful for the task, and
has abundant storage, you have another alternative:
http://static.wikipedia.org/ -- it lacks search, images and
content curation, but it definitely has coverage.

Yes, there's definitely a different set of use cases involved between
shipping gigabytes of HTML on the school server, and shipping a small
copy of the most interesting articles on Wikipedia on every laptop,
standalone.  The main goal of the Wikipedia activities is the latter,
which explains the tradeoffs it makes.

- Chris.
-- 
Chris Ball   c...@laptop.org
One Laptop Per Child
___
Server-devel mailing list
Server-devel@lists.laptop.org
http://lists.laptop.org/listinfo/server-devel


Re: [Server-devel] Schools Wikipedia

2010-03-29 Thread Chris Ball
Hi,

I'm getting 100s of messages on the server when an XO browses to
a new link on the Wikipedia site. They fly by so fast it is hard
to read, but some start with Debug: and the latter ones that
are still on the screen say things like:

Yeah, that's normal; I suggest redirecting stdout and stderr to /dev/null.

- Chris.
-- 
Chris Ball   c...@laptop.org
One Laptop Per Child
___
Server-devel mailing list
Server-devel@lists.laptop.org
http://lists.laptop.org/listinfo/server-devel


Re: [Server-devel] Schools Wikipedia

2010-03-02 Thread Chris Ball
Hi,

The current Wikipedia activities already work as web servers on
the XS, and require much less space due to the content being
compressed -- for Spanish, 25,000 articles and 3,000 images in
100MB.

David pointed out that this isn't actually documented anywhere.  :)

Here's the HOWTO.  On a fresh XS, for English WP:

$ wget http://dev.laptop.org/~cjb/enwiki/WikipediaEN-4.xo
$ unzip WikipediaEN-4.xo 
$ cd WikipediaEN.activity/
$ python server.py en_US_g1g1/en_US_g1g1.xml.bz2 4567

.. then you can browse to http://schoolserver:4567/ on an XO or other
machine.

For Spanish, get http://dev.laptop.org/~cjb/eswiki/Wikipedia-11.xo and
replace en_US_g1g1 with es_PE.

- Chris.
-- 
Chris Ball   c...@laptop.org
One Laptop Per Child
___
Server-devel mailing list
Server-devel@lists.laptop.org
http://lists.laptop.org/listinfo/server-devel


Re: [Server-devel] Schools Wikipedia

2010-03-01 Thread Chris Ball
Hi,

Don’t know if anyone else has tried this, but in the
Schools-Wikipedia (www.schools-wikipedia.org) is a selection of
5,000 articles

I like it. We need good ways to import this (and similar static
content) into an XS. I guess the main use case would be
installing it onto the XS from a USB stick. We also want to make
a nice wiki.laptop.org page with a list of content that can be
used this way...

The current Wikipedia activities already work as web servers on the
XS, and require much less space due to the content being compressed --
for Spanish, 25,000 articles and 3,000 images in 100MB.

Takes longer to render pages than static file content does, though, so
it's a tradeoff between bandwidth and storage space on one hand, and
render speed/CPU use on the other.

- Chris.
-- 
Chris Ball   c...@laptop.org
One Laptop Per Child
___
Server-devel mailing list
Server-devel@lists.laptop.org
http://lists.laptop.org/listinfo/server-devel


Re: [Server-devel] XS-rsync: automatic .contents creation

2009-02-24 Thread Chris Ball
Hi Dan,

As Ties pointed out, it's related to .toc contents files (which
XS-rsync calls .contents) and not content bundles.

The xs-upserv script knows how to create a contents file given an
exploded directory tree -- is that what you're looking for?

http://dev.laptop.org/git?p=users/cjb/xs-rsync;a=summary

- Chris.
-- 
Chris Ball   c...@laptop.org
___
Server-devel mailing list
Server-devel@lists.laptop.org
http://lists.laptop.org/listinfo/server-devel


Re: [Server-devel] Mass olpc-update via XS?

2009-02-18 Thread Chris Ball
Hi Martin,

Interesting. I didn't know cjb had forked my code. Should I fold
his code and ship it with the XS? Chris, can you tell me more about
the patches?

Sorry about that, I should have written this up.

The change here is the addition of an xs-upserv script, and it's
probably not useful for the XS -- you already have an xs-rsync setup
that does the same thing.  We wrote it while in UY for their non-XS
school server, as a quick script to turn any machine into an update
server: it takes a directory tree with a build per directory, creates
an rsyncd.conf that exposes each build, and launches rsyncd with the
new config to serve the builds.

At the time, we needed something to prove that UY's modified 767 build
could be upgraded to by their modified 649 build using their local
school servers, without the assumptions made by the xs-rsync setup.

Thanks,

- Chris.
-- 
Chris Ball   c...@laptop.org
___
Server-devel mailing list
Server-devel@lists.laptop.org
http://lists.laptop.org/listinfo/server-devel