Re: [gentoo-user] Re: Gentoo's future directtion ?

[Alan Mackenzie]

Hello, everybody.

On Sat, Nov 29, 2014 at 07:09:07AM +0000, Martin Vaeth wrote:
> hasufell <hasufell@g.o> wrote:
> > Martin Vaeth:
> >> hasufell <hasufell@g.o> wrote:
> >>>> With rsync I believe you can exclude categories:
> >>>> http://www.gentoo-wiki.info/TIP_Exclude_categories_from_emerge_sync

> >>> That is uninformed.

> >> I think he is right.

> >>> check the --depth option of git. You can even clone specific tags with
> >>> --depth=1.

> >> Every tag will still contain all categories:
> >> AFAIK, with git, it is not possible to update everyting but e.g. *access*
> >> *kde* *i10n* *gnome* if you know that you will never install an
> >> ebuild from these categories.

> > My max DL rate is ~700KiB/s and is the limiting factor.

> My concern is not the time but the total volume (there are still
> often limitations involved), and perhaps even more important,
> the disk usage, especially compared with methods like squashfs(+aufs).
> It simply is a fact that with git you have to download and store a
> lot of unnecessary information (if you are not a developer and do
> not use a heavy system): not only git metadata but also
> unneeded categories.
> So for non-developers, downloading with git does not necessarily
> make sense.

> That being said, please do not consider this as an argument against
> a change to git: For developers it has only advantages, and AFAIK,
> it is not planned to cancel other download methods anyway.

Speaking as a developer in a project which has just converted to git, I
can assure you that git has tremendous disadvantages, even compared with
cvs.

Principally, git does not have a high level model of version control
concepts, so that using git is somewhat analogous to programming in
assembler.  Both give you tremendous control and the ability to do
practically anything, including shooting yourself in the foot.  So that
instead of conceptualising a "branch" (as you would do with Mercurial,
Bazaar, Subversion, or even CVS), you need to think about "commits
reachable from a certain head (excluding commits reachable from some
other head)".

git is very difficult to learn, compared with, say Mercurial.  To
compare, if you do $ git help branch, you get a man page ~180 lines long
dumped on you, and that's taking up the full width of my 240 character
wide screen.  If you do $ hg help branch, you get a 27 line concise
summary, max. ~80 characters wide, which nonetheless is pretty much
complete.

git has become very popular (much as systemd has), possibly because
programmers are frustrated at not being able to write in assembler any
more.  (At least, that's my theory).  Like systemd, it has established a
stranglehold on its domain.  But severe disadvantages it most definitely
has.

-- 
Alan Mackenzie (Nuremberg, Germany).