On Friday 30 March 2007 05:42:51 pm David Brodbeck wrote:
> [EMAIL PROTECTED] wrote:
> > This thread really does show the unfortunate direction that software
> > development has taken even in open source: The simplest package is a rube
> > goldberg-like conglomeration of pre-packaged code and requires 50 and 100
> > other packages, each one recursively depended on it's own set of libs and
> > scripts and packages!!!
>
> That's called "not reinventing the wheel."  Often it's a good thing.


Yes. In terms of software development, it is a matter of philosophy. You 
can "reinvent the wheel" each time or use someone else's code. 

Generally it is better to use what already works.

>
> One example: The zlib bug.  You may remember this one -- a bug in a
> decompression routine that created a security hole.  A lot of packages
> had zlib as a dependency.  While they were all affected by the bug,
> fixing them was just a matter of replacing one shared library.

IIRC, this was pretty widespread.


>
> Other packages simplified things, in the way you suggest, by simply
> including the zlib source code inside their own code.  This meant they
> didn't have zlib as a dependency.  But it also meant that every one of
> these packages had to be tracked down and fixed individually.

Yes, which is more trouble for fixing, and may - or may not - be a good thing. 
I'm honestly on the fence on this idea.

Seeing how really convoluted and complicated the LFS is, I would tend to argue 
for putting libraries in one location.  But that is a different argument.


>
> I'd argue that nine times out of ten, using a pre-packaged library is
> both simpler and more reliable than rolling your own.  It also saves
> space.  Why should every package carry around all the code needed to,
> say, draw a window, when they can link to a single library that does it?
>
> > Creating a darned index should definitely take less time than solving
> > 500,000 equations with 500,000 unknowns about 100 times over, updating
> > the silly thing should be almost instantaneous!!!
>
> Indexing is I/O-heavy, unlike equation-solving.  This isn't a matter of
> CPU power.  Until someone invents a mass storage medium where every
> location can be read instantly, indexing is going to be time-consuming,
> because you have to wait for data to be read off disk.

I'm waiting for the day when I can push my insignia, say, "computer!" and a 
nice lady's voice comes on asking me what I want.

-- 
kai

Free Compean and Ramos
http://www.grassfire.org/142/petition.asp
http://www.perfectreign.com/?q=node/46
-- 
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to