Update of siunitx went wrong (probably)

[George N. White III]

On Wed, Nov 8, 2023 at 8:16 PM Norman Gray <gray at nxg.name> wrote:

>
> Norbert, hello.
>
> If I may jump into this conversation....
>
> Thanks for the very useful explanation.
>
> On 8 Nov 2023, at 23:19, Norbert Preining wrote:
>
> > Which pill do you take?
>
> I can't think of use-cases where the winner wouldn't be:
>
> > - only import packages into TeX Live when (2) AND (3) has happened
> >       possible, not so easy
>

What makes it less than easy?

It would be less confusing if the catalog could include the date the
package version appeared on CTAN or "pending step2".


> >       but many will complain "it is already 3 (THREE!!) days that
> >       the upload is on CTAN but not in TeX Live" ...
>
> I take this is just 'everyone waits for step (5) to happen, with a lag
> of at most a week'.
>
> Self-consistency is a virtue, and this preserves it.]
>
> I suspect that the only people who would even notice a few-day gap
> between the release of an update and its appearance in the archive would
> be those involved with the change, such as someone reporting a bug.  But
> they would presumably have out-of-band access to the updated version.
> Or they could just have some patience: it's an archive, dammit, not
> TikTok.
>

I have little sympathy for people who complain about delays propagating
updates of free software into archives.  If the update is really urgent and
important they should use the "out-of-band" route.

A "pending step2" in the catalog would help the people who are looking for
a particular update to decide whether to wait or update "out-of-band".
There would be one extra step step to replace the "pending flag" with the
date the package actually appeared.


> Also, it seems nice if Karl is permitted a day off at least once in a
> blue moon.
>

The vast majority of users go long periods between installing updates so
only notice delays when a) coauthor is using a newer version generating
different output or noise (logs show different versions), or b) they
encounter a bug and are looking for a newer version of a package.   Some
package distribution systems have schemes where updates go to "testing"
status and need a certain number of positive test reports before the update
is released.  Maybe a slower cadence with "out-of-band" testing could
better handle "urgent" requirements without significant impact on the
majority of users.


-- 
George N. White III
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://tug.org/pipermail/tex-live/attachments/20231108/a9d6ee92/attachment.htm>