On Mon, Nov 19, 2007 at 02:43:00PM +0000, brian m. carlson wrote: > On Mon, Nov 19, 2007 at 02:13:20PM +0100, Werner LEMBERG wrote: > >Robert Thorsby wrote: > >>However, it is truly frustrating for all concerned when The Debian > >>Way turns out to be the wrong way because, as you are aware, The > >>Debian Way does not concede the possibility of error in itself. :-)
For the record, as far as developers are concerned, it generally does concede the possibility of error. It's the more fanatical of our users who tend to be more rigid and inflexible, IME ... :-) The problem with Debian groff is not about some Debian Way - it's simply about trying to avoid regressions in something for which we've promised support. IMO we would have been better off not promising that support in the first place, but it was before my time. Folks have suggested having separate packages. This is technically possible, but it becomes really rather painful to maintain (and tedious to implement in man(1)), and the pain would continue in the form of upgrade difficulties even after it was no longer necessary. Frankly, I've been putting off dealing with the issue rather than having to do that. > >Well, they've solved a particular problem, and the current groff can't > >replace this particular need so I can understand that they don't want > >to upgrade. Do complain more to them: maybe someone at Debian gets > >interested enough (as Bruno Haible was) to add proper CJK support > >based on Unicode. > > I'm happy to (try to) add kinsoku shori handling for Debian[0], if Colin > Watson will tell me what version he wants to update to (1.19.2 or CVS; > the latter is my preference because of UTF-8 input). If he doesn't > say anything, I'm going to assume that CVS groff is acceptable. I would be overjoyed if somebody got us out from between the rock and the hard place. CVS would be fine. Basically the top-level goal is that man(1) keeps on working at least as well as before in Japanese locales. (I think Chinese sort of works at the moment, but probably not perfectly.) If that's taken care of, I'm really not too bothered about how! I am fed up with being stuck at an old version of groff. I was under the impression that groff needed glyph classes in order to implement kinsoku shori. > Do note that Ubuntu has probably not updated to a newer groff because > they probably don't need it, and it would be a regression to CJK support > (which they need) if they did. Note that this is just speculation; I > don't actually know what Ubuntu's needs are, since I only use Debian. My day job is at Canonical (though not speaking for them etc.), and I'm an Ubuntu developer as well as a Debian developer. As such, I am not especially keen on duplicating work for myself by giving them different versions of groff; as far as I'm concerned the two distributions' needs in this area are more or less identical. Thanks, -- Colin Watson [EMAIL PROTECTED]
