Re: Planning: General Timelime
> General Guidelines:
> - Match other Debian architectures as closely as possible, even at the
> expense of being different than the original CygWin distribution.
A few issues with this that we should discuss:
- Distributing Source. It is likely that the diff's required to build some
packages under the CygWin DLL are going to be significant. If at some
point we want to merge the Win32 architecture in to the main Debian
distribution, how can we coordinate these differences with those used
in building for native Linux installations?
- File Upgrades. Upgrading the CygWin DLL is difficult. It cannot be
done while any CygWin executable is running. In fact, any file currently
open cannot be upgrade. This means that all daemons must be totally
stopped during the procedure, but it also makes it difficult to upgrade
the upgrader. How do we deal with that? Perhaps an external, self-
contained program that reads a list of files that must be moved and
does it at the first opportunity, trying every few seconds until there
are no more files to process?
- LIBC. CygWin currently uses "newlib" instead of "glibc". Following
this general guideline means switching that. If somebody has a pointer
to why this choice was made, I for one would appreciate reading it.
- GCC. The CygWin gcc has extra options for supporting native win32
applications. Do we include those as well? I think we probably should
if it's not too difficult and we can deal with the (perhaps extensive)
source-code changes required (see my first point).
- Mount Structure. CygWin uses "/cygdrive" (e.g. "/cygdrive/c") to access
Windows drive letters. Do we continue to follow this? Perhaps we could
simplify it to just "/" (e.g. "/c") followed by a drive letter. Maybe
"/win" (e.g. "/win/c")?
And likely much, much more. What others can people think of?
One could argue that this is a good case for following the CygWin convention,
but I think if that's the case then there is little to be gained from a
separate Debian distribution and thus is not worth the effort to build one.
The CygWin distribution is an excellent one; there is no need to try and
re-invent that wheel.
Brian
( bcwhite@precidia.com )
-------------------------------------------------------------------------------
Differences are good. If two people agree, one of them is redundant.
Reply to: