[Discuss] Debian 12 vs. WSL 1

[grg-webvisible+blu at ai.mit.edu (grg)]

On Thu, Jun 22, 2023 at 02:00:45PM -0500, Derek Martin wrote:
> On Thu, Jun 22, 2023 at 01:43:10PM -0500, Derek Martin wrote:
> > If you're talking about 3rd-party application software, then your
> > points are reasonable

the great bulk of the universe of software installed on *nix machines is
"3rd party" (as opposed to what's copyrighted by the original os "vendor"
as you outlined in your prior paragraph), so that is indeed what the great
bulk of this discussion about developing/porting/maintaining/distributing
software is about (though I personally wouldn't discard the effort of
people just because they're paid by a "vendor" to develop/maintain
software).  as a result, the development, porting, and maintenance burden
of "3rd party" software will be a big driver behind what we are actually
discussing here: the reasons for undertaking usrmerge.

so, thank you for your acknowledgment.

> > but we're not--we're predominantly talking
> > about the OS layers, which includes things like how shell scripts
> > work, the paths at which system utilities lived

"3rd party" software in fact has shell scripts and uses system utilities.

you seem to be discarding anything which hasn't affected you directly,
dismissing rn/trn/rrn out of hand without seeing the bigger picture,
so for some silly reason I'll try again: in the 90s did you use any gnu
tools?  gcc?  bash?  gzip?  emacs?  or vi?  perl?  netscape?  apache?

in that era cross-platform (let's throw in "3rd-party") software typically
had serious ./configure scripts, because they had to, and their developers
and maintainers put in some real effort to make it work on your system.


> infrastructure to address the different paths already existed, because
> it had to--until then, every single Unix vendor had this split

you seem to imply it's automatically handled by "infrastructure" and thus
zero effort for developers and maintainers?  it's not.  it takes human
effort, and the people expending the effort wanted to reduce it via usrmerge.
you got the "had to" part right, but it isn't automatic, soylent green is
people!


> The porting effort that Debian described in their document--as far as
> I can tell--is predominantly to do with the fact that everyone else has
> already undertaken the merge, and so porting from upstream packages is
> more effort on Debian for that reason.  But, if no one had ever
> undertaken the merge, no one else would ever have needed to, and no
> special effort to maintain what already existed would have been required.

it only might seem that way because the effort has always been expended
upstream to make packages compatible with various downstream environments.
upstream of debian is source code; note the explicit mention of ./configure in
	https://wiki.debian.org/UpstreamGuide#Your_Build_System
when the whole world was a scattered mess pre-usrmerge, upstream developers
often put in the work to support at least a couple different downstream
environments, because no single environment was a big enough user base.
if not the original developers, others ported and maintained those ports.
such effort has been expended continually since long before you came on the
scene, whether you recognize or appreciate it or not. your assessment of "no
special effort" really translates to leaving things as they have always been:
those upstream developers/maintainers continuing to expend this "non-special"
effort as they always have.

what you're reading in the debian document is their assessment that if
debian is the lone holdout without a usrmerge-standard environment, some
developers will be satisfied with supporting only the usrmerged
environment, and so the effort formerly expended upstream that debian had
formerly benefited from will now have to be expended within debian.

in short: the upstream has always expended this effort, and got away from
it because of usrmerge.  if debian does not usrmerge, then that effort falls
on debian.  but if debian does usrmerge, then they get away from it as well.

developing, porting, and maintaining cross-platform software has always
required plenty of effort to make it work on your system, and my contention
in this thread has always been that the intended benefit of usrmerge is to
reduce that effort.


respectfully, I feel this conversation isn't advancing; perhaps I'm still
trying to answer a question you had at the beginning ("I'm curious if this
change is thought to have any genuine practical benefit"), but maybe you've
moved on to a different topic ("the problems that Rich and I were discussing,
namely the compatibility and interoperability of different Unix variants")?
in any case I believe I've advanced and explained my singular contention here
detailing the benefit of usrmerge and at this point I don't see that it
would be productive to continue this discussion as it is.  feel free to reply
to get the final word on the record telling me how everything I've said is
wrong and irrelevant and I obviously have no idea what I'm talking about.  ;)

--grg