- From: David Perrell <davidp@earthlink.net>
- Date: Thu, 24 Oct 1996 13:25:04 -0700
- To: "Stuart Young" <nakor@glasswings.com.au>
- Cc: <www-html@w3.org>
Stuart Young wrote: > Unfortunately, the problem is choosing a decent seperator. Dos uses a ., > and many programs on the mac either use a ., _, or a -. It's difficult not being slightly centric to a system you've used for 12 years, however bad you may realize it is. I wanted to pull out of this discussion because (1) I don't feel sufficiently educated about the MacOS and don't have time to get so and (2) I kept edging it toward a stupid Mac/PC debate. And I really don't hate the Mac, even though it never quite suited me. The Mac may be the most elegant desktop system pre-NeXT, but there's something to be said for chaos, too. IMO, the Mac wouldn't be as good as it is today without the explosive hardware evolution on the relatively open Intel/MS side, and NT wouldn't finally be getting a good interface without the Mac. Both platforms have benefited from advantages in the other. But back to file registries. The sensible part of the discussion was simply whether file type should be in machine- or human-readable form. Part of my argument against machine-readable was emotional. Surely the system database does _not_ need to be upgraded in toto from a central database as I suggested -- the applications themselves update the database, just as the file type association database is updated in Win. BUT, aren't there alternatives to a central database? File types have become much less proprietary since the days of .PCX (the worst compression scheme ever invented a 'standard'?! -- damn good thing the PC side stayed chaotic!) and some (Tag Image File Format being especially notable) are structured such that applications can recognize and deal with even enhanced versions of the format, no matter what system the file was created on. Can't universal file types evolve such that a central database is simply not necessary? Why should standardization revolve around a numerical code with no inherent meaning? Version numbers of such file types do _not_ need to be part of a filename extension. As an 'open' file type evolves, programming-related information would be readily accessible. > There is another way. Using such a central registry on the programmers > end, and sticking slightly on some conventions. This means that you have > to binary search the start or end of the file for a particular tag, but > it can work, and isn't that hard to implement. > > The central registry is only useful to programmers to get a list of file > tags so they can embed them in their programs (and since a new file type > would change the tag, you can't use new file types, not that the program > would understand the new file type anyway, because it can't decode the file!) You are vaguely describing TIFF, and a later version of TIFF includes features of the previous. A later TIFF version need not break an application's ability to read the file unless it utilizes tags the application can't understand. Even then some of the information may be usable. It still seems to me that the only need for a central registry of file type codes that are distinct for every version is in deference to the MacOS. And not a good idea. David Perrell
Received on Thursday, 24 October 1996 16:35:37 UTC