Re: Some examples of sniffing usefulness



El lun, 19-01-2004 a las 13:22, Xavier Bestel escribió:
> Hi list,
> 
> first, I'm not there to troll :)
> 
> I'm very happy some measures are taken to bring speed to the otherwise
> nice nautilus, but I was a bit concerned about the removal of automatic
> sniffing. Well, today I stumbled on one of the cases where it's useful:
> I just downloaded a big bunch of .wav files, and some of them were not
> correctly fetched. Even if they were saved as *.wav, they were in fact
> .html files containing some 404 error page.
> There I was pretty happy to still have nautilus 2.4 and to see these
> files stamped with "html" icons. Way easier to find and delete the wrong
> ones !
> 
> So I would understand if some users would prefer the current (old ?)
> behavior. Having wrongly-suffixed files may happen legitimately.

I raise my hand in favor of the old behavior, just like you do.  Exactly
this has happened several times with my setup, and I'm glad it happened
this way =).

The question of using file extensions to identify files is much, much
more than academic, and people advocating their usage as the primary
source for file types for the sake of speed are plainly WRONG.

Correctness over speed, always.  If the file detection code didn't
detect a file properly, the solution is NOT to revert to file
extensions.  The solution to the correctness issue in these cases is to
fix the file detection database.

Now to the speed issue.  Fundamentally, all this speed/extensions drama
happens only because, to this day, Nautilus still doesn't store metadata
on files, despite having mechanisms to do so (the current XML metadata
storage form, and Extended Attributes on systems that support it).  If
Nautilus DID do it, once it had read a large folder, it could save the
MIME type to the metadata store.  Then upon a second visit, the load
would be SO MUCH FASTER (being that both XML files and EAs have much
higher access times than detecting filetypes from their content).

Instead of using the most stupid system ever invented for distinguishing
file types (the ubiquitous and aberrant file extension, its popularity
courtesy of Microsoft), what every file manager developer should be
doing is doing things right: Detecting MIME type once, and saving it in
a previously agreed upon location (which I beg it's EAs) so the next
time it becomes much faster to load it again.  Do it once, do it right,
instead of bending over to established stupidity.

(file extensions are stupid, period.  This has been argued before, very
successfully, by theorists and practical people.  If you want to argue
something about this e-mail, at least pick something good to argue upon,
instead of arguing about file extensions' role in a healthy desktop
experience)

Please look at Robert Love's last e-mail sent to this list (I forwarded
it two weeks ago) before saying EAs cannot be used.  If you read it, and
know who Robert is, you will not be tempted to say they can't - unless
you're a file extension fan, in which case you probably feel insulted by
this message already.

> 
> 	Xav
-- 
	Manuel Amador (Rudd-O)
	GPG key ID: 0xC1033CAD at keyserver.net

Attachment: signature.asc
Description: Esta parte del mensaje =?ISO-8859-1?Q?est=E1?= firmada digitalmente



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]