| 1 |
On Sunday 10 April 2005 8:34 pm, Ciaran McCreesh wrote: |
| 2 |
> On Sun, 10 Apr 2005 20:27:03 +0200 Christian Parpart <trapni@g.o> |
| 3 |
> |
| 4 |
> wrote: |
| 5 |
> | Both have pros and cons. Well, the ASF has everyting converted into a |
| 6 |
> | single repository and they seem to be just lucky with it. KDE is |
| 7 |
> | about to convert everything into a single svn repos as well (for |
| 8 |
> | other reasons). For the Gentoo projects, it might make sense |
| 9 |
> | (administrative) to keep everything into a single repository as well. |
| 10 |
> | However, providing each sub project with its own repository will work |
| 11 |
> | around the single-point-of-failure effect (in worst case) so it's |
| 12 |
> | likely to happen this way. |
| 13 |
> |
| 14 |
> Nothing to do with single points of failure. |
| 15 |
|
| 16 |
maybe wrong said. I mean, when you break the repos, you break |
| 17 |
everything and the whole development process halts. when you |
| 18 |
break a little repos if a single dev group, you break just |
| 19 |
this one (to be fixed though) and the others will continue w/o |
| 20 |
any problems. |
| 21 |
|
| 22 |
> SVN uses transactions and |
| 23 |
> changesets. These make a heck of a lot more sense if they're done on a |
| 24 |
> per project basis. |
| 25 |
|
| 26 |
reason? |
| 27 |
|
| 28 |
> Unlike with CVS, this makes a big difference -- SVN |
| 29 |
> revision IDs are actually meaningful, |
| 30 |
|
| 31 |
SVN repository IDs represent the state of the whole repository at a given |
| 32 |
time, nothing more or less. |
| 33 |
|
| 34 |
> and you don't want to lock every |
| 35 |
> single Gentoo project whilst one person on a slow dialup connection does |
| 36 |
> a single transaction to a single project. |
| 37 |
|
| 38 |
as confirmed by svn devs and others, the transaction data is first uploaded to |
| 39 |
the server (with whatever speed the client has) and then performed |
| 40 |
server-side. Though, the time of locking the database depends on the CPU |
| 41 |
load, and not the client's [dialup] speed. |
| 42 |
|
| 43 |
Hmm... besides, the ASF is just having a single repository for all their |
| 44 |
public projects (with about 1000+ contributors) w/o any problems. |
| 45 |
|
| 46 |
Regards, |
| 47 |
Christian Parpart. |
| 48 |
|
| 49 |
-- |
| 50 |
Netiquette: http://www.ietf.org/rfc/rfc1855.txt |
| 51 |
23:44:22 up 18 days, 12:50, 2 users, load average: 0.51, 0.64, 0.72 |
Archives