ZPublisher publishing retry policy should be configurable

[Chris McDonough]

Currently, Zope retries a request at most exactly 3 times before giving up. The need to retry a request is always? due to a ConflictError. In some high-write-concurrency non-performance-critical applications, it would be useful to allow administrators to be able to set this number higher. We should un-hard-code this policy and make it configurable via a config file option.

[Andreas Jung (ajung)]

Would be interesting for Zope 2.8. On the other hand I don't know how important this issue will be if we have MVCC in Zope 2.8

[Chris McDonough (chrism-plope)]

MVCC only prevents ReadConflictErrors, normal ConflictErrors (which happen at commit time) still exist.

[Ben Liles (bliles)]

3 years and no progress on this?

[Benji York (benji)]

Ben Liles wrote:
> 3 years and no progress on this?

Your patch must have gotten lost.

[Ben Liles (bliles)]

Here is a patch for this that works with Zope 2.10.5.

[Ben Liles (bliles)]

Oops, wrong patch submitted earlier.

[Ben Liles (bliles)]

I have attached a patch that takes care of this.

[Andreas Jung (ajung)]

No ides why this bug is in stage "fix released".

The patch does not apply the current trunk and I should have a test.

[Tres Seaver (tseaver)]

Note that long-term, we mean to be able to run Zope by default under a WSGI server, where this feature can be handled by middleware such as repoze.retry:

  http://pypi.python.org/pypi/repoze.retry

In the near term, I will merge your patch for the Zope 2.12 branch and the trunk.