> Some users may have downloaded 2.6.3 while it was on Maven central. If > you now release 2.6.3 you could end up in the position where if a user > reports they are using 2.6.3 you don't know which of the two 2.6.3 > versions they are using. > > If the diff between the accidentally released RC and what ends up being > formally released as 2.6.3 has zero functional impact then you are > probably OK but there always the risk that there will be something. > > It is worth considering throwing away that version number and moving to > 2.6.4 for the next release.
Normally this is a reasonable point to consider, but in this case, actually we don't have to worry about that. A fact I didn’t explain in depth in previous email is that the unexpected artifacts can not be successfully resolved by users, because it missed the most basic part 'dubbo-parent’. So I think it’s ok to continue with 2.6.3, you can see the latest vote thread which I’ve already started in here: https://lists.apache.org/thread.html/fdbb88c98d149c7bcefe9709316350c1d1f95463fad3e6148c63c05d@%3Cdev.dubbo.apache.org%3E <https://lists.apache.org/thread.html/fdbb88c98d149c7bcefe9709316350c1d1f95463fad3e6148c63c05d@%3Cdev.dubbo.apache.org%3E> > More generally, version numbers are viewed as 'cheap' at the ASF. It > isn't a big deal to decide not to use one for some reason. Tomcat, for > example, doesn't use RCs. Most of our release votes pass first time but > if they don't we fix the issue, increment the version number and try again. > > I have seen other projects decide not to use a version number of various > reasons that all, generally, boil down to confusion over exactly what > that version number represents. Moving to a new version number is often > the simplest way to avoid potential confusion. Good to know that, a new perspective on version number for me to learn. I have always been thinking that every open source project must try to avoid skipping versions, I think they only do that when having no choice but to increase the version number. Now, I can see some open source projects not act that strictly with continuous version numbers, and they work well with that. Best regards, Jun > On 28 Aug 2018, at 22:06, Mark Thomas <[email protected]> wrote: > > On 28/08/18 05:56, Huxing Zhang wrote: >> On Mon, Aug 27, 2018 at 10:44 AM jun liu <[email protected]> wrote: >>> >>> The sonatype team have deleted all artifacts of 2.6.3 from central >>> repository. I will start 2.6.3 RC4 vote now. > > Just a thought. > > Some users may have downloaded 2.6.3 while it was on Maven central. If > you now release 2.6.3 you could end up in the position where if a user > reports they are using 2.6.3 you don't know which of the two 2.6.3 > versions they are using. > > If the diff between the accidentally released RC and what ends up being > formally released as 2.6.3 has zero functional impact then you are > probably OK but there always the risk that there will be something. > > It is worth considering throwing away that version number and moving to > 2.6.4 for the next release. > > More generally, version numbers are viewed as 'cheap' at the ASF. It > isn't a big deal to decide not to use one for some reason. Tomcat, for > example, doesn't use RCs. Most of our release votes pass first time but > if they don't we fix the issue, increment the version number and try again. > > I have seen other projects decide not to use a version number of various > reasons that all, generally, boil down to confusion over exactly what > that version number represents. Moving to a new version number is often > the simplest way to avoid potential confusion. > > Mark
