From: jvs@netcom.com (Jonathan Stockley) Subject: Re: [NEW]: The Linux Device List Date: 8 May 1993 00:16:14 GMT
In article <1993May7.110338.22901@mnemosyne.cs.du.edu> ijackson@nyx.cs.du.edu (Ian Jackson) writes:
>In article <1993May5.102510.12491@mnemosyne.cs.du.edu> I proposed:
>
> tty[a-d] for the `standard' 4 ports
> tty[e-o][0-f] for multi-port cards (ie tty<card><port>)
>
>I stand by the above, despite Rick Miller's contrary proposal in his
>latest release of the Device List.
Isn't this partly to do with what this whole games is about?
If you don't like the way something is done, you have the right AND
capability to change it.
>
>IMHO the scheme with tty[S..]n is ugly, too small to fit the range of
Why is ttySn ugly as a serial port but not as a pseudo tty? :-) :-)
>cards and ports in without crowding and making the namespace hard to
>fathom, interferes with part of the most natural extension of the pty
>namespace, is unnecessarily complicated for the most common cases, and
>uses unnecessary capitals.
>
>The only advantage I can think of is that it is an extension of what
>it seems some people have been already using, namely ttyS[0-4] for the
>standard ports - though there seems to be some confusion, so this
>could be considered a disadvantage.
I agree there was some confusion when it was change from ttys[0-4] to
ttyS[0-4]. For example Xconfig had the mouse on ttys0. But wont there be
just as much confusion if we switch to tty[a-d]?
>
>In article <jvsC6KIqw.4GC@netcom.com> Jonathan Stockley answered my
>>> - similar to a major commerical Unix (ScumOS), and thus familiar
>>> to most Unix people.
>with
>>I question the implication that *most* Unix people are using ScumOS. Do you
>>have any figures to back this up? Don't forget the non-workstation world.
>
>I didn't say that most Unix people are using ScumOS; I just said that
>most would be familiar with this scheme. It is true that I may be
>overestimating the breadth of experience of some Unix people, though.
No you IMPLIED it. :-)
>
>>Hmm, I have never heard of this "accepted convention". I have looked at an
>>SVR3 system that uses [pt]ty[p-z][0-f] first then [pt]ty[a-o][0-f] next when
>>the first lot have all been used.
>
>Interesting; obviously it is not an accepted convention everywhere - I
>stand corrected. However is it really relevant ? Are you proposing
>the scheme you describe ? It seems to me that tty[p-zP-Z] is better,
>firstly because it brings the pty namespace closer together, and
>secondly because it allows us to use tty[a-o] for the serial ports as
>I propose, which is better in other respects besides freeing tty[P-Z]
>for ptys - see my original posting.
>
>> On the SvR4 systems I looked at they have
>>/dev/pts/0, /dev/pts/1 on through /dev/pts/511 [...]
>
>Hopefully you're not suggesting such a scheme for Linux.
No, just throwing it in for contrast :-)
I think that Rick's proposal is fine and should be used as the "standard"
way to go for distributions such as SLS, Yggdrasil et al. The fact that it is
at least documented and maintained and published here gives it more than a
little weight, surely?
If the MAKEDEV script in /dev made them this way then those people that
prefered a different scheme could simply edit that script.
One last point before I meander home, why no discussion about other
device names?
Jo
-- Jo Stockley jvs@netcom.com jo@88open.org