On Sat, 2003-11-15 at 01:05, Stefano Mazzocchi wrote: > On 15 Nov 2003, at 00:47, Oliver Zeigermann wrote: > > > Stefano Mazzocchi wrote: > >> I think users should be able to do > >> cvs co jakarta-slide > >> ant > >> ./slide > >> and get it running. The required (non-optional) jars should be > >> included in the download or fetched by the build script from jar > >> repositories (the only problem seems to be JTA which is under the Sun > >> license, so we can't put it in CVS, but the Geronimo guys already > >> reimplemented it under the apache license, so we can use that). > > > > I disagree. If we finally have a release why would any *user* want to > > do this? > > because they download the binary distribution, then find a bug, then > checkout the CVS module, then build. > > If the build is not piece of cake, those guys will simply not continue > and you loose yet another potential committer. Look around: you'll see > that the simplicity of CVS building is directly proportional to the > number of active committers that that project has. It's a cause, not an > effect. > > > On the other side: I have seen so much branching and even painfully > > merged part of it myself, partly because people seem to be afraid to > > contribute their stuff publicly. If people get the impression > > everything that is in the CVS HEAD needs to be in production quality > > valuable contributions might be deterred. > > Nah, slide is highly modular, you can plugin your new stuff without > breaking anything (as you are doing with the new stores). > > At the same time, if you change something that is core, this *must* be > kept in production quality mode. that is: you should always do a clean > build and pass the unit tests before committing. > > using strict checking for core and looser checking for pluggable stuff > will give you solidity and ease of innovation, without sacrificing one > for the other. >
This sounds highly reasonable! > -- > Stefano.
signature.asc
Description: This is a digitally signed message part
