Question about version in submitting patches

[Pascal Bourguignon]

Hi!

Le 27/11/2020 à 13:51, Marco Antoniotti a écrit :
> Hi
>
> Sorry for the general noise, not necessarily related to the issue at hand.
>
> I know I am a P.I.T.A.,  but I kind of concluded that versions of the kind
>
>     YYYYMMDD
>
> Are better than
>
>     major.minor.small.itsy.bitsy.bit
>
> What do you think?


Well, using a timestamp as version number is not as informative for the 
user as the semantic major.minor.bug version number.

The usual meaning being that:

- the major is incremented when incompatible changes to the API are 
made: users updating from one major to another should expect to have to 
invest some work to upgrade their stuff for the new version.

- the minor is incremented when compatible changes to the API are made 
(additions to the API, or change, with a compatibility layer provided): 
users updating from one minor to another can expect to only have to 
recompile their stuff for the new version, if at all, and no more work.

- the bug is incremented when bug corrections are made, without any 
change to the API: user updating from one bug to another can expect 
total compatibility and no work at all on their part.


Instead of that, if you use a timestamp as version number, you now have 
to keep metadata, such as what versions are LTS (long term support), or 
other such attributes.  This works for whole distributions, but not for 
single libraries.


-- 
__Pascal Bourguignon__