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

file placement, redux



Bruce Sass <bsass@freenet.edmonton.ab.ca> writes:
> > You forget the extension.
> 
> Nope, I left them off.  ;)

We need to leave the .bin extensions, I think.  One reason is just
that people are accustomed to seeing extensions denote the file type.
Two, it's convention with boot-floppies (ain't broke, don't fix).
Lastly, we servers ought to have the extension defined as
application/octet-stream.  They usually leave default as text/plain.

> Technically, yes, safe is a special, but what would you call it if it
> was only one manufacturer's box that required the safe option...

Safe is not really a special, nor is it a subarch.  Special and
subarches are actually quite similar.  Basically, anything where we
need to have a different set of boot and drivers, paired, is either a
subarch or a special.  Safe is neither -- it's just alternative boot
code.  So I'd prefer to just have, for i386:

  disks-i386/disks-1.44/rescue.bin
                        rescue-s.bin

(as it is now).  Dignifying this with a directory is too much.  Maybe
at most:

  disks-i386/disks-1.44/safe-opt/rescue.bin

Then I could put the README-safe readme in there...!

> What it boils down to is that a sub-set of boxes out there have special
> requirements; whether those requirements arise because the mfg hung
> different chips off the bus, uses a different OS, or a slightly broken
> floppy drive (this is safe, as in the `safe, slow and stupid' floppy
> driver option, right) is not important.

Yes -- it kinda is.  Replicating a whole dir tree just for a frob of a
value of a boot-loader, or designing around such a nusicence, is going
to confuse users.  At the top level, you need to present as few
choices as possible.

> So, for all practical purposes, safe can be treated like a sub-arch
> (but only because the i386 platform is the only one that needs safe,
> clearly this would not work if every arch under alpha also had a
> safe version).  Plus, doing otherwise would throw a wrench into my
> attempt at automating the generation of url entities for the docs.

I don't see how this is so for safe....  It is just something that
deserves special mention for i386 only.  Again, we don't wanna design
the whole system around it.

> Having more directories to drop things into could mean that less
> information needs to be encoded into the filename;
> would that help with 8.3 compatibility?

Yes, that was the point....

-- 
.....Adam Di Carlo....adam@onShore.com.....<URL:http://www.onShore.com/>


Reply to: