Thanks, finally an answer about the process - which you ignored last week :)
I personally don't see why we need to move the site from svn to git as a
pre-requisite, but if someone is doing that, great.
Now, in general I agree, but I feel it a little bit strange that you
allow a new git repository for atomos but want to block everyone else. I
suggested the three modules, where I see active development; we have
open issues for them, regularly get feedback etc. In how far are these
different?
And of course, if a project gets moved it requires that the history is
kept and tags are moved etc. but there are existing scripts out there
which do this.
That said, I'm fine to wait but then I want to see that people are
tackling the open points you mention. I'm happy to help where I can of
course.
Regards
Carsten
On 02.03.2020 07:52, Karl Pauls wrote:
Wait, I am objecting :-).
In the first place, while I realize there are benefits to having a git
project matching a subproject (and yes, I would like to move framework
at one point), most I can think of correspond with active development
or at a minimum, need work beyond just creating the repo and doing a
move - hence, I think we should only consider such a move if at least
one out of these two are likely for a given subproject (preferably,
both).
That in part is why I was ok with atomos because, it clearly has some
active development atm and is running a lot of tests and builds in
actions etc (the other part is Tom really insisting). For SCR,
configadmin, and http I could accept the second argument (mostly
because they have big-ish test suites) - provided we have the
commitment to not only move them but really set them up so that PRs
get checked automatically, the readme make sense, etc. I have a harder
time seeing it for fileinstall and utils atm mostly because there has
been very little activity or we don't have that much testing or
documentation in place. The framework, I have quite some things queued
up in my sandbox for connect - so I guess some level of active
development applies but I would still call it a bit of a stretch atm
unless we'd really set-up the test suite too.
In the second place, I strongly feel we should define a process (incl.
all necessary steps/actions) for module breakout and get our house in
order first. I don't think it is a good idea to start breaking out
modules without a plan. It is unclear how we would break out modules
(what are the steps/scripts to keep the tags, etc.) and how is that
explained/workable e.g., how would you find/check out all modules,
etc. Furthermore, we didn't yet update the website with the current
change to our source repo (let alone, migrate the website itself to
git). Starting to break out arbitrary modules is not going to make it
more likely to have that documented and explained in a good way.
Having two (and a read-only, almost empty) repos is one thing to
convey and work with - tens (or like some other project have, even
hundreds) is another. Yes, all of theses things are solvable (and
solved in one way or the other by other projects) but we need them
solved for us and I suggest doing that first (including consensus).
In other words, in general I'm against to breaking out modules until
the following is addressed:
- the website is migrated from svn (and cms.apache.org) to git and
something that works with asf.yaml
- the website content is updated to reflect the current git move
(instructions, locations, links to svn, the works)
- we have an agreed upon set of steps/actions/scripts to use to break
out a module (incl. history, tags, website updates, etc).
- and a documented way of working with our repos (how to find, check
out together, etc).
Furthermore, wrt. "Separate git projects for SCR, Configadmin and
http" until we have a proposal for the improvements we would get by
breaking them out and have volunteers to implement them.
Finally, I'm against "other modules" until we determined on a
case-by-case vote basis that we assume a development activity uptick
is imminent and somebody proposes (and volunteers to implement)
reasonable improvements we would get by breaking them out.
That said, I don't think any of the above is super hard to achieve -
hence, if you (or others) are volunteering: let's get it done. Maybe
start with creating JIRAs for the above and start discussions around
how to achieve them?
regards,
Karl
On Mon, Mar 2, 2020 at 7:24 AM Jean-Baptiste Onofre <[email protected]> wrote:
OK, I will start with these repos.
I keep you posted.
Regards
JB
Le 1 mars 2020 à 14:26, Carsten Ziegeler <[email protected]> a écrit :
I think moving to a separate git makes sense for active (in some sense)
projects. So, yes, I guess it makes sense for the three you mentioned, too.
+1
Carsten
On 01.03.2020 12:41, Jean-Baptiste Onofre wrote:
+1 for that.
Do you want me to tackle other modules (I’m thinking about fileinstall, utils,
framework to start with) ?
Regards
JB
Le 1 mars 2020 à 12:11, Carsten Ziegeler <[email protected]> a écrit :
Hi,
I would like to move the SCR implementation, Configadmin implementation and
http implementation into separate git project each (only one for all http sub
projects).
If no one objects, I'll go ahead in the next days.
Thanks
Carsten
--
Carsten Ziegeler
Adobe Research Switzerland
[email protected]
--
--
Carsten Ziegeler
Adobe Research Switzerland
[email protected]
--
--
Carsten Ziegeler
Adobe Research Switzerland
[email protected]