Hi.

Le lun. 12 juil. 2021 à 19:42, Rob Tompkins <chtom...@gmail.com> a écrit :
>
> +1
>
> Note. The number of jars for the number of classes available seems a bit 
> extreme. It also makes it quite difficult to validate the release given that 
> to properly validate the release (note, proper validation involves validating 
> all of the signatures of all of the files generated both in nexus and in 
> subversion) one has to traverse 324 files across 14 different urls. It took 
> me upwards of 1.5 hours to validate this properly.

As I've pointed out several times, it doesn't seem that this
particular step needs
to be done "manually": A script could be generated that automatically performs
the following actions:
 * downloads all the files
 * downloads the checksums files
 * runs the appropriate checksum validator
 * creates a report of all checks, to be validated by the reviewer.

That this is not done yet should not preclude sane modularization (i.e. subject
matter based).
[Going the other route was tried and it failed:  By ever growing Commons Math,
it became unmanageable, in several ways (not going to restate the details once
more, it's all in the ML archive): Not a single release in more than 6
years despite
a very long list of issues, is proof enough IMHO.]

Also let's not forget that the official release is made of source code.
Most of the 324 files which you mention are convenience binaries compiled and
uploaded through a tools chain which we generally trust.

>
> All the best, and good work,

Thanks,
Gilles

> > [...]

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org
For additional commands, e-mail: dev-h...@commons.apache.org

Reply via email to