[Nut-upsdev] NUT protocol versions (was Re: [nut-commits] svn commit r3618 - in trunk: . docs server)

Charles Lepple clepple at gmail.com
Wed May 23 12:15:18 UTC 2012


On May 23, 2012, at 4:58 AM, Arnaud Quette wrote:

> Hey Charles,
> 
> 2012/5/23 Charles Lepple <clepple at gmail.com>:
>> On May 22, 2012, at 7:47 PM, Charles Lepple wrote:
>> 
>>> On May 22, 2012, at 10:49 AM, Arnaud Quette wrote:
>>> 
>>>> Author: aquette
>>>> Date: Tue May 22 14:49:15 2012
>>>> New Revision: 3618
>>>> URL: http://trac.networkupstools.org/projects/nut/changeset/3618
>>>> 
>>>> Log:
>>>> Add versioning to the network protocol
>>>> 
>>>> Due to the number of recent evolutions to the network protocol, along
>>>> with other possible future changes, NUT network protocol is now
>>>> versioned. The chapter "Network protocol information", of the developer
>>>> guide, stores the revision history, while a new "NETVER" command serves
>>>> this information through the protocol
>>> 
>>> 
>>> I really thought we had discussed not using numeric version numbers :-(
>>> 
>>> This should really be more of a "does the server have this capability?" question.
>> 
>> 
>> Perhaps "discussed" is an exaggeration... "mentioned":
> 
> damn, it's part of the threads with Greg, that I completely missed,
> and that I still have to actually read...
> and since Greg is still not back, I've left these aside for the time
> being, to focus on 2.6.4.
> 
> the NETVER mod is more a basic way to finally track the protocol changes.
> I was hesitating between an introspection system, that indeed allow to
> check the capabilities, and a simple "version" that implies these
> caps.
> 
> so, looking again at the mail you pointed, I'm still unsure if I
> should revert NETVER, or leave it since it could still be an
> interesting basic check, WRT an introspection mechanism...


I'd recommend reverting, but we've had the discussion before about putting too much meaning into version numbers (when assigning features to a release) and the protocol version number seems similar: the arguments for and against aren't particularly strong.

The only case where introspection might not be helpful is if you want to query for a capability, but don't want to actually run a command yet. For something like the EPO or DYING status, this might be useful.

-- 
Charles Lepple
clepple at gmail






More information about the Nut-upsdev mailing list