[tech] A question of File System

[David Cake]

At 3:48 PM -0700 23/4/02, Ian McKellar scribbled:
>On Tue, 2002-04-23 at 13:29, David Cake wrote:
>>  >  >
>>  >>  And if you want pretty pink flowers, well, I'm sure Apple will have
>>  >>  a filesystem with them soon... in five fruity flavours.
>>  >
>>  >Dom (ex-Be guy who wrote BFS) is working on a new FS at apple. That
>>  >should kick ass. BFS had all sorts of really nice things like
>>  >journalling and indexed file attributes. If Apple let him implement that
>>  >kind of thing it'll be great. The lack of attributes and the lack of
>>  >automatic metadata indexing on unix filesystems sucks. If OS X got that
>>  >it just might make me give up OS 9.
>>
>>	Its about time. A good filesystem has been an obvious lack of
>>  Mac OS X for some time, and there was much rejoicing in some circles
>>  when Dom was hired.
>
>A friend of mine (a long-time NeXTStep user) tried running OS X with
>UFS. It became clear very quickly that Apple just don't test that :-/
>>

	And plenty of open source projects that should know better 
assume all the worlds case sensitive, too.

>  > At 2:20 PM +0800 23/4/02, Nick Bannon wrote:
>>  >That's definitely got some promise. It's kinda irritating at the moment
>>  >because it doesn't have the simple ext2 attributes (immutable bit,
>>  >etc), but there is or has been a patch to provide that. What it ought
>>  >to get at some stage is more generic extended attributes, eg somewhere
>>  >to store ACL's for Samba.
>>
>>	Not to mention handling those wacky Apple file attributes
>>  (Type, Creator, etc) properly.
>>	One of the particularly tragic things about Mac OS X is that
>>  old school NeXT people are always getting all worked up about HFS+s
>>  use of file attributes etc, and advocate using Apples fairly shocking
>>  UFS implementation instead, seemingly completely unaware of the sort
>>  of file features that have made it into modern unices.
>
>Yeah, some of my friends at Apple are frustrated by a lot of that. Some
>of the NeXT people don't want to have inode monitoring (like linux's
>dnotify/imon/fam, freebsd's kqueue, irix's imon/fam, BeOS had it, WinNT
>has it...). I think my housemate (who works on the OS X display server)
>is just gonna hack it into Darwin :)

	Some of the NeXT guys seem stuck in a bit of a time warp as 
regards Unix stuff. Hopefully Jordan Hubbard will set them right. At 
least he will hopefully make them remove all those man pages that 
were last updated in 1993 and document things that no longer ship 
with Darwin.

>A couple of my friend who work on the Finder were at Be, one (Pavel) was
>the guy who did most of the work on the Tracker (their file manager)
>which exploited extended attributes extensively. Once they get that the
>Finder might kick ass.

	Tracker did it right. A similar system would be a good thing. 
I'm glad to hear some clueful people are working in the area.

>Unfortunately Apple seems to move very slowly
>with technology (based on hearing things from people who work there and
>having to wait years to see it ship).

	Yes. If you follow some of the Darwin things closely you can 
see it take 6 months or more from 'finished' to 'ship' sometimes.
	A lot of Apple seems to currently be stuck in a warring 
tribes mentality. And the current management of software don't seem 
to be fixing it*. In at least one case I have satisfied myself that 
this is because they are an idiot :-)

	Cheers
		David

*excluding Bud Tribble, who hasn't been at Apple long, and might turn 
out to be good.