minor correction, it should not read "sullify" but pacify. It has been a long week ....
Just my two cents, I like a clear indication, that I am working with a development version beyond the obvious changes of text. SO the versioning is usually the first thing I look at.
That being said, I could imagine adopting the Python versioning scheme as an alternative to the current even/odd numbering.
As for the number of branches - I allways thought, that actively maintaining that many branches is more hassle than it's worth. I know it happend now due to special circumstances, and I like a clear indication of break (aka the main Qt reasoning behing the 2.8<->3.0 switch), but I think it presents more confusion then help to the customers and just tends to sullify the little nerds in all of us ;-)
kind regards
Roland
We currently have three active release branches: 3.0, 2.6, and 2.4. This is because we support each release branch for a set amount of time (typically 24 months after the initial .0 release) and our last three .0 releases were less than 12 months apart. However, having many active branches can sometimes cause confusion[1] and far fewer people download the "Old Old Stable" release than the "Old Stable" or "Stable" releases. Would it make sense to have only two release branches active at any given time, e.g. by adjusting our release branch lifetimes to "24 months or whenever we have two newer active branches, whichever comes first"?
I think two active release branches makes sense, but I'm not sure that it always makes sense to have them be the two newest stable releases. When people decide what release to download (or when Linux distributions build a package, since we want to consider not the direct download stats but also what gets bundled with distributions) the primary consideration is the necessary libraries, not the features. Some stable release branches add a lot of Wireshark feature but don't add a lot of library requirements. For example, I think all that 2.6 added over 2.4 was requiring a version of CMake that all distributions still in long term support already had. Since it's difficult to find a system that could install 2.4 that couldn't also install 2.6, it's really not necessary to support 2.6 and 2.4 simultaneously (and that would explain the lack of 2.4 downloads.)
OTOH, 3.0 bumps a lot of library versions, so if 3.2 (or whatever it's called) is relatively minor in requirements changes (but heavy enough in features to justify a new branch), I could see it making sense to keep 3.2 and 2.6 around, and drop 3.0 support first.
Cheers,
John Thacker
Sent via: Wireshark-dev mailing list <wireshark-dev@xxxxxxxxxxxxx>
Archives: https://www.wireshark.org/lists/wireshark-dev
Unsubscribe: https://www.wireshark.org/mailman/options/wireshark-dev
mailto:wireshark-dev-request@xxxxxxxxxxxxx?subject=unsubscribe
