Hi Gwern, Let's have another look at this.
On Fri, Nov 21, 2008 at 16:46:42 -0500, Gwern Branwen wrote > Eric, now that you're project lead, do you want to reconsider Dr. > Roundy's decision to refuse this patch? Let me preface this by saying that I am committed to us spinning off darcs code into third party libraries, some ideas for which being listed on this page: http://wiki.darcs.net/index.html/DarcsLibraries But I also agree with David that we should be very careful not to silently break compatibility with old formats and old repositories. On Tue, Aug 19, 2008 at 10:05 AM, David Roundy wrote: (quoted by Gwern) > > As I mentioned before when asked about this, I don't think it's a good > > idea. No reason to add the extra complexity for a function that we > > only use to maintain compatibility with old repository formats. The > > best way to keep compatibility with old formats is to not change the > > code. So before making a decision on this, I would at least like to see some research done on the point. What are we replacing exactly? What is this function for, and why does David say its only use is to maintain compatibility with old repository formats? What are the risks? Are they worth taking? (Feel free to just point me to past discussion) > Now that we're cabalized, we could use flags. (Or just make > utf8-string a depended-upon package period, I suppose, which would let > us remove src/UTF8.lhs?) The approach I would favour to spinning this off is with a highly controlled phasing out: - first rewrite our code to imitate new APIs - then add ifdefs with the default being our old code (yuck; this is things getting worse before they get better) - then switch the defaults - then jettison the old code (with great rejoicing) And to make sure that we keep moving forward on these phases, we should probably add some notes in a sensible place when we can move forward to each phase. Of course, it doesn't have to be this way exactly (maybe what I propose has serious flaws, like making it unlikely that the new code be tested thoroughly). Folks like Duncan can give us better advice on how to go about this process. But the point is that we should be very careful and make sure we do things in a rigorous and well-controlled way. And if things still break, then at least we can say that we took every effort to keep everything working. Anyway, I am very keen to have you working with us on this! But I would ask for you to show great patience with our ponderous way of working. We try to be more conservative than most projects, because the perceived cost of failure is quite high. Whether or not we actually succeed in preventing these failures, I think our users would at least appreciate attempt at scrupulous care. P.S.: You may argue that this conservatism is ultimately not in our users' best interest. I would welcome a debate on that point, although maybe in its own thread... -- Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow> PGP Key ID: 08AC04F9
pgpn7olAgczF8.pgp
Description: PGP signature
_______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
