El jue, 29-01-2004 a las 06:40, Olaf Frączyk escribió:
> On Tue, 2004-01-27 at 09:32, raphael bosshard slm admin ch wrote:
> > Not only did I try reading the debate, I even think I understand most of it.
> >
> > As far I understand the main concern against extended attributes is based in
> > the idea that reading them is slow. While this may be true ( And I am not
> > sure if this statement is true. As far as I know, EAs are cached. Some more
> > information anyone?) it is de...
> Why do you think it is faster? The slowest thing is to open a file for
> sniffing. The sniffing itself is not so slow.
While sniffing the file requires the file to be opened, it is expected
that small metadata such as EAs (because they are frequently read) can
be in cache, and be faster than reading parts of a file.
Also while we're at it, once EAs are adopted, it is expected that a wide
variety of metadata be stored along them (perhaps physically near), so
reading metadata from EAs will be a one-time operation, instead of the
current algorithm, which basically scrobbles info from different parts
of the disk (.desktop files, hidden homedir files, stuff). Hence, once
the MIME has been read, other metadata will be in the cache (because
linux has a readahead, and most disks also have it).
It also solves another problem: the file access time doesn't change
whenever you use EAs to determine the mime type. Nowadays, opening a
folder with Nautilus updates the file access timestamps for every file
in the folder. Slower. Write access.
(please don't tell people to mount noatime. I myself use noatime, but
there are several valid uses for atimes, and you can't tell users of
atime to just mount noatime)
> About the same delay
> should be for accessing EA. BTW., there are other problems like
> permissions etc.
> And this what I wrote above was already discussed on the list.
> >
> > The funny part is, that even Microsoft knows, that "file extensions" are a
> > bad solution. They try to hide them ("Do not show extensions of know file
> > types") and, by doing so, create some kind of a pseudo-attribute but without
> > any clear concept. Let`s not do the same mistake they did some twenty years
> > ago.
> The biggest mistake is to hide these extensions. If they are shown, then
> you see blabla.doc.pif if they are hidden you see blabla.doc. Nothing
> more to say.
Sorry Olaf. From any perspective you see it, the file extension should
NOT mean anything special. Using part of the file name to define a file
type created one of the most complicated conundrums in computing. The
thing is, your brain is just used to them. It shouldn't. But you've
lived all your life looking through green glasses, and you don't know
blue and red. That's okay. But now that the glasses are off, please
inform yourself on why it is wrong to use file extensions.
(file extensions could still be used during a short transition phase).
>
> Regards,
>
> Olaf
--
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