Hi, And thank you for advice and remarks. I'll tell you when I'll find good solution, if any :-) bye Neil Graham <neil@cs.toronto.edu> writes: > Hi Boris, > > On Sat, 26 Oct 2002, Boris Daix wrote: > >> I feel the solution is not so far... I can say >> that, for sure, the software you mentionned can export UML works in >> XML-like formats, that's already a not-so-bad thing. > > So if I understand your plan, you're planning to figure out the > (indirect) mapping from UML to XMI, then use an SGML editor to write XMI > documents that you'll then feed to a UML generator. Evidently, this is > indeed possible. > >> Well, psgml is an emacs-mode, not a parser. But anyway, I've read >> that SGML parser can read XML, as HTML : the Python modules for SGML >> are often used to parse HTML, as it's a "tag-fashion" language too. > > HTML is an SGML language; XML is not. The trouble with using an SGML > editor to generate XML is that you're very likely to generate an > ill-formed XML document, which will cause an XML processor--like the one > that must underlie the UML generator--to barf all over your shoes. I'm > certain emacs will have some kind of native XML mode, and equally that > emacspeak will support it well (TV Raman helped develop VoiceXML, so I > daresay he'll have completely solved this problem. :) ) But this is > an anthill compared with the mountain of the core problem... > >> He he... I believe that if I say to my teatchers "Hey, look, with >> GNU/Linux, I'm able to work with UML (via XML)", I'm sure Open Software >> would be clapped for hours ! :-) > > No question. If I were you though, I think I'd unearth my manual tactile > diagram-drawing tools and crank UML out that way. That's how I did > digital circuit diagrams way back when and it worked well enough; but then > everyone else was working manually too, so my disadvantage wasn't > acute. But I'd analogize the XMI-to-UML solution to writing Java by using > a binary editor to produce bytecode, then using a disassembler to induce > Java from it... In fact, that'd probably be much easier since bytecode > maps naturally back to Java, whereas XMI only maps to UML via the indirect > route of the MOF. :) > > Anyway, best of luck! I dunno about other folks, but I'd love to hear how > you end up solving this problem; it's one more and more of us will face in > the future, no doubt about it. > > Cheers, > Neil > > > > _______________________________________________ > > Blinux-list@redhat.com > https://listman.redhat.com/mailman/listinfo/blinux-list > -- Boris Daix "Feel free to be Free, or not to be..." _______________________________________________ Blinux-list@redhat.com https://listman.redhat.com/mailman/listinfo/blinux-list