Fwd: 'open' instead of 'xdg-open' for usability?
simon.mcvittie at collabora.co.uk
Mon Dec 16 08:02:59 PST 2013
On 16/12/13 15:21, Robert Qualls wrote:
> The fact that one-word command names are such a limited resource
> means that they shouldn't be used by obscure programs.
openvt is probably older than some of the people on this mailing list
(its git history starts in 2007, but the man page is dated July 1996),
and at the time it was written, it was probably no more obscure than
anything else in the Unix toolbox. Unless you have a time machine,
it's far, far, far too late to object to its former name.
Whether you like it or not, there is no central authority that names
software, and no way to prevent people from naming their software
badly (unless we empower governments to do so, or something, but I
don't think anyone really wants that). The closest thing we have is
that distributions (most visibly Debian, but I'm sure "app stores"
also do this a lot) sometimes refuse to include a piece of software in
their distribution because its name is too generic, or because it
collides with something they already have. See the extensive mailing
list threads regarding Node.js in Debian for an example that happened
> I think we need to have a group of people sit down and hammer out
> something before we get too far into the future. Like, there should
> be some kind of POSIX Command Standard or something that keeps
> track of reserved keywords that do obvious things.
And then what? How does this group impose their world view on the
other 99.99% of the world?
POSIX is not a forced standard
cannot unilaterally impose requirements on Unix distributors.
freedesktop.org is even less like a forced standard than POSIX: it can
document things that the major desktop environments agree on, but has
no way to coerce any of them.
François Revol wrote:
> Btw, debian already has something do deal with this issue
> (although not exactly meant for this one but rather competing
> programs for the same purpose):
Alternatives are specifically *not* for this purpose: they are for the
situation where running "generic-name --options args" (for at least
some subset of the possible options/arguments) does the same thing
whether generic-name points to one implementation or another.
For instance, if openvt didn't have the "open" alias, and the generic
name "open" was managed by alternatives, it'd be OK if xdg-open and
gvfs-open were registered as implementations of "open"; you could type
"open http://blahblahblah" (interactively, or in a script) and have
more or less the same thing happen.
It would not be OK for an "open" alternative to have openvt and xdg-open
as implementations - if you have a script which runs "open", either it
wants to open a new virtual console, or it wants to open a file. Either
way, getting the other meaning is not acceptable.
We can never fix that for open(1), because it's too late. Authors of
new software can avoid it by namespacing generic names, which is why
we have things like gnome-disks and kde4-menu; or by using non-generic
"brand names", which is why we have Evince, Firefox and Amarok instead
of "document viewer", "web browser" and "audio player".
More information about the xdg