Re: [hybi] Issue tracker

[SM <[e-mail]>]

Hi Ian,
At 12:57 AM 5/13/2010, Ian Hickson wrote:
>If there's N issues, and it takes P time to walk each one through the
>proposed procedure plus Q time for the editor to do the edit, then it will
>take N*P+N*Q time to address the issues. However, if instead we just
>do the edits and only go through the procedure for controversial issues
>(issues where the change results in a request for a reversion of the
>change or where the resulting change is not satisfactory), and we assume
>that there's a fraction F of issues that are controversial, then it would
>take N*P+N*F*(Q+P) time.

First of all, it is important to keep in mind that an editor 
expresses the consensus of a WG.  We could take a procedural approach 
where every change goes through what you described above.  Or we can 
have a less stringent approach where we do editorial changes, go 
ahead with uncontroversial changes and tackle the rest as issues.  If 
these issues require a revision of the I-D, we end up with I-Ds that 
do not reflect the rough consensus.  If an implementation is tracking 
the I-Ds, this can cause problems.

Although reversions may happen in practice, it is problematic 
process-wise if we assume consensus and require a reversion of the 
change when someone raises an issue.

>This means it's more efficient to use this "escalation" process rather
>than putting everything through the heavy-weight process if F is less
>than 1/(P/Q+1), which, assuming the fraction of controversial issues is
>less than 80% and the time for the editor to edit the spec is on average
>60 minutes, is true for any value of P greater than about 4 hours, which
>seems quite low given the proposed process.

The IETF process is not light-weight.  The composition of the WG 
influences the speed of the work.  If the editor has to wait for the 
outcome of consensus calls to resolve each issue, the person is going 
to spend much more time editing the specification.  And that could 
happen with an "escalation" process.  If the editor does not try to 
reflect the consensus of the WG, people may end up being annoyed and 
take any action they see fit.  The end-result will be more than inefficient.

Regards,
-sm