[Gllug] Median cut algorithm

[Nix]

On 17 Mar 2008, Adrian McMenamin uttered the following:
> I guess because I've never had such messages from anyone that I've always
> assumed nobody else is using it.

I've written several programs that have had single-digit messages from
anyone, but which I *know* are in use by at least hundreds of people,
because they're elisp packages, and I can see the feature symbols
popping up in (X)Emacs bug reports. (One of them even had a patch! one
single patch! in 2004...)

If your package works and is general enough that new features aren't
often needed, few people will patch it, or comment on it.

> At least now I have a reason to continue work on it. I stopped last year
> when it did what I wanted it to do.

Well, work on it when you need to. Otherwise, leave it be until someone
else hacks it :) it's not as though software rots...

> Main thing is that it is slow (at least when it comes to rendering stuff
> on the web). Currently I am using it to render 30 PNGs in parallel on a
> web page and that is fine when you have a single user (takes about 10

Well, that's much more than I'm doing with it :) it gets a few rendering
jobs every half hour in my application.

> seconds to render the lot), but when the page was slashdotted last year
> the usuage count soared to over 100 and people reported the page as
> non-rendering when in fact what was wrong was that the perl scripts were
> taking for ever.

munin's graphing suffers from this problem too. You can generate the
graphs when the data is updated, which eats a minute of CPU time every
five minutes, or you can do it on the fly when the page is viewed, which
takes a minute or three whenever anyone looks at the stats. Neither is
exactly ideal...

-- 
`The rest is a tale of post and counter-post.' --- Ian Rawlings
                                                   describes USENET
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug