[RFC] Proposal to change the structure of the website view/download areas and handling of "testing" releases

Kevin P. Fleming kpfleming at linuxfromscratch.org
Mon May 17 10:06:15 PDT 2004


Jeroen Coumans wrote:

> I understand your concerns, but in practice it has always worked well in 
> our previous development model. We can let it run this way for some 
> time, see if it still works out; I'll tweak the labels on the website to 
> reflect that they are daily renderings.

OK, I think that's a reasonable course of action. If significant 
problems occur because of it then it can be left that way indefinitely.

> If you don't want to run the risk, I propose to remove the testing & 
> unstable download & read links from the website so they are only 
> available by reading development.html. This is basically your earlier 
> proposal to make the location "unavailable" to the LFS community. 
> Developers & interested users can still use the daily rendering, while 
> more innocent users who heed the warnings of labels marked "testing" and 
> "unstable" are protected.

No, I don't think that will be necessary. Really I'm just trying to 
ensure that we can minimize the amount of time that "testing" requires 
to move to "stable". In the past it's been (apparently) quite difficult 
to get enough solid test reports from the _identical_ build instructions 
for the editors to feel that they could move on... witness the recent 
glibc tarball upgrade just before 5.1 was released: the reports on 
lfs-dev were all over the place (good, bad, bad on certain platforms, 
unsure) and some of that appeared to be because the reporters were not 
following the identical steps to do their builds. In my mind these 
reports only serve to delay the testing process, because they makes the 
editors concerned that there are problems, when those problems are not 
actually present. Moving to a model where people are testing only "dated 
testing snapshots" seemed (to me) to be an ideal way to ensure that 
people could test a known quantity and be able to report _exactly_ what 
build instructions they tested against.

With all that said though, if we do start to create "dated testing 
snapshots", render them and put up news items about them, then people 
can still do most of what I was hoping for. We just need to get someone 
(Matt?) to decide if this is the right policy to follow for the next 
round of testing and decree that :-)



More information about the lfs-dev mailing list