I'd like the time to ask for likely impact on user@accumulo. -- Sean On Apr 15, 2015 5:18 PM, "Christopher" <[email protected]> wrote:
> It seems you may have read my comment about *not* duplicating as > carefully as I initially read your request to comment on the JIRA :) > If you want, I can copy it over there (or somebody else can), but > right now, I don't think it's necessary. :) > > Sorry if I'm adding to the confusion. > > -- > Christopher L Tubbs II > http://gravatar.com/ctubbsii > > > On Wed, Apr 15, 2015 at 4:40 PM, Josh Elser <[email protected]> wrote: > > No worries, thanks for copying it over there. > > > > > > Christopher wrote: > >> > >> Oops, sorry, I overlooked your request to comment on the issue and > >> commented here instead. I won't duplicate my comments there, but will > >> watch both threads for following up. Sorry. > >> > >> -- > >> Christopher L Tubbs II > >> http://gravatar.com/ctubbsii > >> > >> > >> On Wed, Apr 15, 2015 at 4:21 PM, Christopher<[email protected]> > wrote: > >>> > >>> I've been in favor of deprecating mock entirely. I certainly don't > >>> like it being considered public API, since it mainly was built to > >>> enable testing (where heavy flexibility is needed to test various > >>> scenarios... much more flexibility than a stable public API can easily > >>> provide). I also don't like the idea of Shell's Java API being > >>> included in the public API (it's intended as a command line utility, > >>> not a programmatic interface to Accumulo). > >>> > >>> Given all that, along with the inherent difficulty of preserving > >>> MockShell in the public API, I'm thinking this might be one of those > >>> "ask for forgiveness" cases, and make a note about it in the release > >>> notes. I'm just speculating here, but I think this is probably low > >>> risk and won't affect too many people. > >>> > >>> As we think about API stabilization issues moving forward, it's > >>> probably inevitable we're going to run into this occassional case. > >>> What's important is that they become fewer and farther between, and > >>> that we document clearly when it happens. > >>> > >>> -- > >>> Christopher L Tubbs II > >>> http://gravatar.com/ctubbsii > >>> > >>> > >>> On Wed, Apr 15, 2015 at 3:02 PM, Josh Elser<[email protected]> > wrote: > >>>> > >>>> In case others don't see it in JIRA notifications, here's some broad > >>>> visibility. > >>>> > >>>> I noticed, through japi reports, that MockShell was technically public > >>>> API > >>>> and was removed from it due to the addition of the shell maven module. > >>>> If > >>>> you have opinions, please raise them on ACCUMULO-3726. If we need to > >>>> expand > >>>> it out to a standalone conversation, we can start a new discussion > here. > >>>> > >>>> -------- Original Message -------- > >>>> Subject: [jira] [Commented] (ACCUMULO-3726) Restore MockShell in > public > >>>> API > >>>> Date: Wed, 15 Apr 2015 18:59:59 +0000 (UTC) > >>>> From: Josh Elser (JIRA)<[email protected]> > >>>> Reply-To: [email protected] > >>>> To: [email protected] > >>>> > >>>> > >>>> [ > >>>> > >>>> > https://issues.apache.org/jira/browse/ACCUMULO-3726?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14496703#comment-14496703 > >>>> ] > >>>> > >>>> Josh Elser commented on ACCUMULO-3726: > >>>> -------------------------------------- > >>>> > >>>> _curse_ > >>>> > >>>> I was worried about that. This seems like > >>>> yet-another-public-api-shouldnt-be-public-api issue. MockShell came > from > >>>> o.a.a.c.client.mock.MockShell and went to o.a.a.shell.mock.MockShell. > >>>> Shell > >>>> came from o.a.a.core.util.Shell and went to o.a.a.shell.Shell. It > makes > >>>> 0 > >>>> sense to assume MockShell is public API but Shell is not (and also > >>>> causes > >>>> more problems as it leaks Shell into the public API). > >>>> > >>>> Given that, I would be in favor of breaking our rules on > compatibility. > >>>> > >>>> 1. MockShell should never have been in public API (as Shell was never > >>>> public > >>>> API) > >>>> 2. It is not easily restored and deprecated. > >>>> 3. It is likely very low risk that someone was using themselves (we > >>>> ourselves only reference it in a single test case). > >>>> > >>>> cc/ [~busbey] since you have been very adamant in the past on topics > of > >>>> this > >>>> matter. > >>>> > >>>>> Restore MockShell in public API > >>>>> ------------------------------- > >>>>> > >>>>> Key: ACCUMULO-3726 > >>>>> URL: > >>>>> https://issues.apache.org/jira/browse/ACCUMULO-3726 > >>>>> Project: Accumulo > >>>>> Issue Type: Sub-task > >>>>> Components: client > >>>>> Reporter: Josh Elser > >>>>> Assignee: Josh Elser > >>>>> Priority: Blocker > >>>>> Fix For: 1.7.0 > >>>>> > >>>>> > >>>>> MockShell was moved from o.a.a.core.client.mock. to o.a.a.shell.mock > >>>>> which > >>>>> is a violation of our public API statement. > >>>>> It needs to be restored and deprecated with a proper cycle. > >>>> > >>>> > >>>> > >>>> > >>>> -- > >>>> This message was sent by Atlassian JIRA > >>>> (v6.3.4#6332) >
