Are you *sure* we can't put a narky message when iPhone users click a
video? Adobe do!
http://twitpic.com/kf361
(assuming it's real - can anyone with an iPhone please check?)
- d.
How does wikimedia handle starting and stopping lsearchd daemon?
Is it handled by things like /etc/init.d/lsearchd?
Could you distribute /etc/init.d/lsearchd along with lucene-search package?
I have written an extension for embedding Google Waves on a wiki page:
http://www.mediawiki.org/wiki/Extension:GoogleWave
In one of my test wikis this is working like a charm:
http://mickenordin.se/~micke/phase3/index.php/Huvudsida
on another it is not:
http://mickenordin.se/wiki/Wave
On the problem wiki it looks like the javascript isn't loaded properly and I
need to use &action=purge to get the javascript to load at all.
Same problem occurs in a vanilla install of 1.16alpha (r57355)
Why is this do you think? What can I do to get the javascript to get loaded on
each page view?
/Micke
Hello to fellow MediaWiki hackers,
I'm writing to let you know that Yaron Koren and finally decoded to organize
New York MediaWiki Users Group:
http://www.meetup.com/MediaWiki-New-York/
Events are going to be dedicated to various uses of MediaWiki on public web
sites as well as enterprise.
If you have a conference room or some other space in New York and you're
willing to host some of our events, let me know!
If you're in New York or passing by at the right time, feel free to just
join the group!
Thank you,
Sergey
--
Sergey Chernyshev
http://www.sergeychernyshev.com/
I thought that not all attachment types are removed...
---------- Прослеђена порука ----------
Тема: A new functionality
Датум: субота 3. октобар 2009.
Од: yoosef pooranvary <yoosefp(a)gmail.com>
За: smolensk(a)eunet.yu
Dear Mr. Nikola Smolenski
I am sending you this letter in order to ask you to add a new
functionality to the Poem extension which you created for Mediawiki
application. I tried to add this functionality on my own but according
to my unfamiliarity with Mediawiki API, I couldn’t go any step
further. So, I’ll write down the specefics of this functionality:
Poems in Persian, Arabic and some other languages have rime and two
lines are written in one line. Note that I am from Persia (Iran) and I
speak Persian, but I try to suggest a general rule for this
functionality.
There are many types of poems in Persian language which are mentioned
below, together with their rime patterns (They have identical
counterparts in Arabic):
Ghazal:
A A
B A
C A
D A
E
A
Ghete:
A B
C B
D B
E B
Masnavi:
A A
B B
C C
D D
E E
Mostazad:
_________A ___A
_________A ___A
_________B ___B
_________A ___A
_________C ___C
_________A ___A
_________D ___D
_________A ___A
Robaee:
A A
B A
Dobeyti:
A A
B A
Or
A A
A A
Cheharpareh:
A B
C B
Tarjiband:
A A
B A
C A
D A
E
E
F F
G F
H F
I J
K
K
Tarkibband:
A A
B A
C A
D A
E
E
F F
G F
H F
I J
E
E
Mosammat:
A A
A A
B
C C
C C
D
Probably it is not important for you to understand the exact type of
these poems and as you may have found out that lots of them are like
each other.
Code Specification
I ask you to add a “Bilinear” parameter to the Poem extension so if a
user wants to type a Persian or Arabic poem, he just types <poem
bilinear> above his poem and ends it with </poem> tag. The user will
write his poem one line in a line but your code should display two
lines in one. (I call this, Pairing Operation!)
Another issue which remains unresolved is that some lines should be
one in a line and above each other (you may have noticed them in the
above patterns). I suppose that you add an operator like vertical line
(|) to your code. If a user puts this character before a line, that
line should be displayed in the middle without its pair line. The user
should add another vertical line to the next pair line so that line
becomes in the middle and exactly under its pair line. Pairing
operation should restart from the next line again.
Another request is that you add another parameter with the name of
“Gap” so one may use it to choose the size of the blank spaces between
pair lines in a line. For example if a user types <poem Bilinear
gap=10> your code should put 10 “ ” between all of the pair
lines. Of course this parameter should have a default value and I
suggest it to be 10.
The user should also be able to choose the place of his poem in the
page. He should be able to choose from “Right”, “Center” and “Left”.
But the default should be “Center”.
Therefore the final pattern of the poem tag should look like this:
<tag-name, bilinear, gap-space, place></tag-name>
Example:
<poem bilinear gap=’15’ center>
(The poem text)
</poem>
Using gap and center should be optional.
An important thing to remember is that in each pair the right one
should be aligned to the right and the left one should be aligned to
the left.
Please inform me whether you will work on this project or not. Also I
have a Mediawiki installed on my computer and I will be happy to test
your code at any stage.
Please consider adding my Wikipedia username (Yoosef Pooranvary) in
your code if you want. I am waiting to receive your response ASAP.
Truly Yours,
Yoosef Pooranvary
Wikipedia Contributor
yoosefp(a)gmail.com
-------------------------------------------------------
[Resending mail; the first copy I sent apparently did not make it to the list]
Hi,
apologies for the long mail. Here is a summary:
We are experiencing overly high memory needs on some of our MediaWiki
websites. I tried to trace the problem but my practical findings leave me
somewhat puzzled. In particular, I initially expected Semantic MediaWiki to
cause some of the problem, but now it seems that the problems are rather
caused by the fact that SMW can easily generate long outputs, not by SMW's
memory need for creating these outputs. I am not getting any further with
debugging the problem, and I do not know how to improve SMW/MW to avoid it.
== How to do memory profiling? ==
I tried to enable PHP memory profiling in xdebug, but all I got was time data,
and I gave up on this for now. The aggregated outputs in profileinfo.php were
not very useful for me either; in particular, I think that they do not take
garbage collection into account, i.e. they only show new memory allocations,
but not the freeing of old memory. So one piece of code may allocate 20M but
never need more than 4M at a time, while another consumes the same amount and
keeps it due to some mem leak. Especially, the sums and percentages do
apparently not show the real impact that a piece of code has on using PHP's
memory.
So I based my memory estimations on the minimal PHP memory limit that would
not return a blank page when creating a page preview
(ini_set('memory_limit',...);). This measure is rather coarse for debugging,
but it might be the one number that matters most to the user. The results were
reproducible.
== Sudden memory explosion and other findings ==
The strangest observation I made on my local machine (Kubuntu Linux, no caches
whatsoever, MW 1.16alpha r56781, PHP 5.2.6-3ubuntu4.2) was that there tends to
be a sharp boundary between "no memory problem" and "massive memory
consumption". Even long pages could be generated with as little as 4M of PHP
memory. However, if they got just a little too long (one additional line in a
table), then only a memory limit of 50M or more was sufficient to get a
result. I disabled extensions for the test. The table I used for testing was
about 62K with roughly 150 lines, and it used HTML tags, CSS classes, and MW
links.
The findings really indicate a kind of explosion, not a gradual increase.
Disabling extensions increased the maximal size of the table by a few rows
each, but the explosion still happened in the same way. Of course, it is not
clear how useful the PHP memory limit method of measurement is here.
== What to do about it? ==
How can I fix this? I noticed that it helps to shorten the input (e.g. I can
render longer tables if the CSS class names in the tables are shorter!), but
also if I simplify the input (replacing links by plain texts). The table I use
is based on HTML syntax: I did not try yet if MW's pipe syntax leads to better
performance. So one option would be to try and simplify SMW's table code
(maybe making it less readable).
But this would only shift the problem towards longer tables. Caching certainly
would also reduce memory consumption, but I have observed very high memory
need on sites even with APC as a PHP bytecode cache (as I said: loading less
code simply moved the problem by a few lines). I did not try object caching
(memcached or APC) yet -- is it expected to help here? Squid does not solve
the problem, since the page still needs to be rendered at some point, and the
memory limit must be high enough to allow this (and if there is a cache miss
and rebuilding the cache takes very long, then there can be another cache miss
while the first request is not done rebuilding the page -- we have had this
killing one of our servers recently).
Besides this, I would still like to know which measures in SMW or MW could
help to reduce the problem at its source. Maybe it would help to know which
aspects of parsing a table are having the highest impact on MW's memory usage.
Or is this a PHP issue?
Any relevant insights/pointers are welcome,
Markus
--
Markus Krötzsch <markus(a)semantic-mediawiki.org>
* Personal page: http://korrekt.org
* Semantic MediaWiki: http://semantic-mediawiki.org
* Semantic Web textbook: http://semantic-web-book.org
--
I got this suggestion for new functionality in Poem Extension. I don't think I
even fully understand it, so please consider if it could and should be done.
Hi,
I was wondering if anyone here knows how often does Google update
their Wikipedia layer for Google Earth and how do they get the data? I
mean, they use the geo templates in the articles, but do they parse a
wikipedia dump or they use a crawler or some other method.
I'm not sure if Wikimedia has anything to do with it, but I think I
have a better chance of getting an answer here than by asking Google
(the company) directly. Google (the search engine) was not really
helpful on the matter.
Thanks,
Andrei