Re: DEVFSv50 and /dev/fb? (or /dev/fb/? ???)

Anthony Barbachan (barbacha@trill.cis.fordham.edu)
Thu, 6 Aug 1998 23:28:49 -0400


>
> Your analogy would work if one were comparing an older version of
>devfs to the current version. But if you want to liken devfs to the
>most recent C++ standard, the current /dev scheme is much more like C.
>And as we all know, people still program in C.
>

I do not agree with you here, at least from what I understand from the FAQ.

>
> /dev/sd? may be simpler than /dev/scsi/c0t0u0d0, but it is essentially
>arbitrary -- if you add a SCSI device with an ID between two others, then
>half your fstab is likely to break, since /dev/sd? is assigned in the order
>the devices are scanned. I wouldn't call that very clean at all. As for
>IDE, I would be inclined to agree -- but there are EIDE CD changers in
>existence, which argues at least for the need for /dev/ide/cXtXuX.

EIDE CD changers might be better addressed with /dev/hda[1-x]. Simple and
eliminates the need for more devicenames.

>
>[snip]
>
>> I think the verbosely cryptic naming sceme of the current devfs is the
only
>> real problem with it. My suggestion is to simplify the naming (keeping
>> backward compatability when possible).
>
> It's not really that cryptic. It's verbose, yes, but not moreso than
>is necessary -- I don't like the fact that adding devices to my SCSI chain
>can require me to boot off of a floppy because all of a sudden my root
>device isn't what it used to be and Linux can't mount its root fs.
>

It is cryptic. I can be pritty sure about that when most of the UNIX gurus
I know have to double check vfstab or /dev/dsk when doing something to a
Solaris device. Might be needed but still cryptic. And sd[a-p] could be
hardcoded to SCSI ID's. Or something similar.

>-- Michael
>

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html