On Tue, Mar 31, 2020 at 13:08:05 -0400,
Matthew Miller <mattdm@fedoraproject.org> wrote:
We did communicate as the very top line of our gathered requirements that
open source is essential to our community and central to our feedback. I'm
not trying to be soft on that. Let's just not do purity-test level
assessments and instead focus on our goals.
The response from CPE made it sound like they just counted requirements rather than evalutating how important each requirement was to each group. Perhaps that was not intended, but that's they way it sounds. I think that being able to theorectically switch from hosted to self-hosted in short order (like in a month), should have been a deal breaking requirement from Fedora in case something at Gitlab changed that prevented using their hosted service. That implies having access to the source (capable of running our instance) with a free license and regular exports of the data in our hands. It doesn't sound like that is a requirement from the response provided by CPE.
Because of switching costs, this is likely to prevent us from going back to Pagure if it does develop a vibrant independent community. That would be unfortunate.
This particular dilemma reminds me very much of the time when the
LInux kernel developers weren't using version control, and it
became clear that one is needed. Linus just refused to use CVS,
and after some controversy, the core developers decided to use
Larry McVoy's proprietary BitKeeper distributed VCS. This solved
the technical problem and was successfully used for few years
(2002 to 2005, IIRC).
Bitkeeper critics were pointing out that while the Linux
community was free to use it to keep the source code, the
Bitkeeper terms of use prohibited the non-commercial users from
extracting the metadata (history, logs, etc). This issue kept
causing problems, finally spurring Linus to sit down and invent
git, and the rest is history.
It is important to remember that BitKeeper, while proprietary,
had a very friendly and close relationship with the FOSS
community, both when they joined forces and even when they were
parting ways. Still, the official Linux git log (
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/?ofs=905000
) lacks the development history preceding 2005-04-16, and starts
with one heck of a commit:
2005-04-16 | [PATCH] mmtimer build fix | Christoph Lameter | 1 | -1/+1 |
2005-04-16 | Linux-2.6.12-rc2v2.6.12-rc2 | Linus Torvalds | 17291 | -0/+6718755 |
Perhaps the relevant lesson is that the only permanent thing is
that nothing is permanent. Decisions that seem inevitable and
superior do not necessarily continue to be so, and it's good to
have a contingency plan for such event----although I am pretty
sure that Linus did NOT plan to work on git in 2002.
Disclaimer: I wrote down my personal best recollections and
opinions. Please draw your own conclusions and analogies; I ask
for your indulgence hoping that it will be enlightening to at
least someone.