[ntp:hackers] ntp release numbers

Martin Burnicki martin.burnicki at meinberg.de
Mon Oct 17 07:58:41 PDT 2005


Hi all,

While the discussion here is mainly about the release version numbers,
I'd like to add some thoughts about the version numbering of ntp-dev.

Also, normal users are unable to tell whether a ntpd version with a
higher patch level number is newer than a version with a lower patch level.


Brad Knowles wrote:
> At 4:23 AM +0000 2005-10-12, Harlan Stenn wrote:
>>  Because we do not currently "publish" anything except the ntp-stable
>>  release and then any snapshot tarballs from ntp-stable.  There is no
>>  easy way to show if folks are running the released ntp-stable or one of
>>  the follow-on snapshot releases of ntp-stable.
> 
>     Fair enough.  I can see that we need some sort of mechanism of
> determining precisely which version of ntp-release/ntp-stable someone is
> running, and I think we should differentiate this from ntp-dev.

Absolutely.

We have seen several times in the near past that the patch level numbers
in ntp-dev have been completely messed up after some repos had been
merged (BTW, I'm assuming that this depends only on which repo is merged
into which other repo).

This makes it nearly impossible to keep track of the changes or relate
any versions/patch level numbers to comments in bugzilla.

For example I'd like to see something like:
"bug x first occurred in version y and has been fixed in version z."

>     If we change ntp-stable relatively infrequently, then incrementing
> the third digit should be sufficient.  If we change it frequently, I'm
> not sure what would work best.  And I'm not sure precisely what we
> should take to be "infrequently" versus "frequently".  Is ten times a
> year too often?  Twenty?  I dunno.

Agreed that the 3 digit should be sufficient for the stable branch.

>     But I'd like to see an ISO date-based scheme for ntp-dev.

Also agreed.

>     I'm thinking of something similar to the FreeBSD or postfix release
> numbering schemes, but not quite exactly the same as either. Maybe X.Y.0
> would always be an official "ntp-release" version, whereas X.Y.Z (where
> Z>0) would be an ntp-stable update, and X.Y-yyyymmdd would be an ntp-dev
> snapshot.
> 
>     People would be able to tell at a glance which version you were
> running, and would be able to easily determine when that version was
> created, and therefore which code was or was not included.

The name of the ntp-dev tarballs currently include the date when the
tarball jhas been generated. The tarballs also include a "version" file
which I assume is generated when the tarball is prepared (normally when
"make" is run).

I'd propose to either add an additional file to the tarball which
contains the date of the tarball creation, or to put the date into the
"version" file that is shipped with the tarballs. Maybe the BK
patchlevel might also be kept, so that the version would be composed of
the base version:

4.2.0b (for now)

and an extension with or without the patchlevel, e.g

20051017
30051017-1.1417

This way we could als give a unambiguous names to the GUI installers for
Windows.


Martin
-- 
Martin Burnicki

Meinberg Funkuhren
Bad Pyrmont
Germany


More information about the hackers mailing list