The only thing to consider here is that having a site separated from repo with docs never really worked and it's ALWAYS out of sync. Inviato da Yahoo Mail su Android Il ven, 19 gen, 2018 alle 2:00, Paul Gale<paul.n.g...@gmail.com> ha scritto: >I generally agree that the documentation should be part of the code otherwise it is out of alignment.
If by 'alignment' you mean that the doc is correct with regard to the source it shipped with can be inferred because they came from the same repo/commit? If so that doesn't make sense to me. Ensuring alignment, as it were, requires additional correlating metadata which does not require that they both live in the same repo. When they're in different repos that affords one the ability to have separate and distinct management strategies for each, e.g., being able to relax the requirement of PRs for documentation. Thanks, Paul On Thu, Jan 18, 2018 at 6:32 PM, Owain McGuire <owain@integration.technology> wrote: > Paul, > > One of the hardest aspects of using Camel is the ability to read the > documentation. We generally use “pair-reading” to interpret unfamiliar areas > - reviewing the examples, tests and source code are more productive. I > generally agree that the documentation should be part of the code otherwise > it is out of alignment. However, a wiki ethos seems more alive to me. Pull > requests seem too heavy as you say. > > Are we really talking about the need for a User Guide Wiki, somewhere between > adoc and the “books”. Is there a place for a Cookbook of example routes with > pointers to the “hard” docs to aid adoption? > > In the back of my mind I think there could be something like a library of > “lego routes” which are working routes on Github that use components to do > real life examples using components and EIPs. e.g. pick up a message and > send to slack, pull an order from Amazon Marketplace, raise an order in > Netsuite. The exchange architecture of properties and headers which are > propagated provide a mechanism to have very late binding on route endpoints. > Morph into RAAS (Routes as a Service)? TNBT?? Perhaps a step too far at the > moment but something to counter the “clicks v code” objections I get from > customers. How far has Mulesoft pushed this? > > O. > > >> On 18 Jan 2018, at 22:44, Paul Gale <paul.n.g...@gmail.com> wrote: >> >> Trust me, I have no love for Confluence as a product. However, even >> with only editor rights I could work completely autonomously when it >> came to editing the documentation. The ease with which I could update >> the documentation made me all the more willing to do so, even for >> simple typos and reformatting, nevermind correcting horrific grammar. >> To have that same ability in the new scheme would require committer >> rights. Going through a pull-request process for documentation is >> antithetical to the egalitarian nature of a wiki. The responsiveness >> of committers to accepting a pull requests is beside the point; if >> they're accepting said requests by reflex then they're not adding any >> value so why have them in the loop. Now, however, I will hesitate >> before considering an edit which more often than not will result in it >> being abandoned - it's just human nature. However, on the ActiveMQ >> wiki, for example, I rewrote/reformatted/reworked whole pages that >> were out of date, on a few occasions spending days doing so. Such >> large scale edits, in contrast to a minor correction, increase the >> likelihood of either push back or rejection if they have to submitted >> through an approval process. The irony of such changes being approved >> by the very people whose initial offering is being reworked is not >> lost on me. Any errors in content I may/will make can easily be fixed >> by another editor making a correction of their own. That's a more >> effective way to work than a gated approval process. All part of the >> wiki concept I suppose. >> >> Regardless, this situation can be avoided if the documentation is >> managed like a wiki and made open to the general public for editing. >> Can that be done? >> >> Thanks, >> Paul >> >> >> On Thu, Jan 18, 2018 at 5:09 PM, Brett Meyer <br...@3riverdev.com> wrote: >>> Pull requests are still a thing, right? ;) >>> >>> Kidding aside, the GitHub pull request and review process seems highly >>> preferable to fighting the wonky Confluence platform, especially since it >>> gives the community a chance to chat about proposed changes before they're >>> merged. >>> >>> What about that is concerning? And with the exception of the eventual >>> merge/commit, why would that limit the documentation pool to committers >>> only? From my experience, the Camel committer team have always been >>> incredibly quick to respond to and work through pull requests... >>> >>> >>> >>> On 1/18/18 5:00 PM, Paul Gale wrote: >>>> >>>> Brett, >>>> >>>> Thanks for your response. >>>> >>>> You have confirmed my worst fears about the documentation solution. Oh >>>> well, all those future edits I had in mind, gone. >>>> Thanks, >>>> Paul >>>> >>>> >>>> On Thu, Jan 18, 2018 at 4:55 PM, Brett Meyer <br...@3riverdev.com> wrote: >>>>> >>>>> Hey Paul, I asked the same question a couple of weeks ago -- Claus >>>>> reminded >>>>> me about the move to asciidoc in the central repo: >>>>> https://github.com/apache/camel/tree/master/docs/user-manual/en >>>>> >>>>> However, we might consider at least adding a note to the tops of the >>>>> current >>>>> Confluence docs (assuming there's a way to do that without editing every >>>>> single page) mentioning the stale state and future plans. Better yet, >>>>> could >>>>> we consider setting up a redirect sooner rather than later, even if >>>>> that's >>>>> temporarily to GitHub (maybe >>>>> >>>>> https://github.com/apache/camel/blob/master/docs/user-manual/en/SUMMARY.md)? >>>>> >>>>> >>>>> >>>>> On 1/18/18 4:34 PM, Paul Gale wrote: >>>>>> >>>>>> Can someone with the relevant privileges investigate why snippets >>>>>> being referenced by the Camel wiki appear to be universally broken and >>>>>> what can be done to fix them? They are an integral part of the >>>>>> documentation and need to work. At a glance I can see that most >>>>>> snippets reference source files from an SVN repository which would >>>>>> explain the breakage. However, I don't know where they should point >>>>>> instead as removing the word 'trunk' in the file path doesn't fix it. >>>>>> It would seem that snippet support is no longer available - I don't >>>>>> see any reference to them in the Atlassian documentation. Perhaps said >>>>>> support came from a plugin that's no longer installed? Guessing. Any >>>>>> info about that would also be appreciated. >>>>>> >>>>>> I understand that the current Confluence backed wiki is generated via >>>>>> some scheduled process. Can that process itself be documented and >>>>>> access to it granted to the entire community? I would have thought >>>>>> opening up access would increase the likelihood of it getting fixed. >>>>>> I've edited a number of pages on both the ActiveMQ and Camel wikis, >>>>>> however I am not a committer (and have no plans to become one) and >>>>>> therefore cannot step in to fix it when it breaks. >>>>>> >>>>>> I recall hearing plans that Confluence would be replaced by some other >>>>>> documentation, perhaps a wiki on Github (ugh). What's the latest on >>>>>> that front? >>>>>> >>>>>> I do hope that whatever solution is settled on does not require one to >>>>>> be an Apache committer in order to edit the wiki documentation. Such a >>>>>> requirement would be unacceptable to me. However, it seems to be >>>>>> likely based on the solutions I've heard proposed elsewhere (assuming >>>>>> the Camel community follows suite with the ActiveMQ community who >>>>>> appear set on such an approach - reasonable assumption given the >>>>>> overlap between the respective communities) that documentation be >>>>>> embedded in the sourcecode, extracted using a tool that would then >>>>>> generate the site, or something similar. Any approach along those >>>>>> lines would likely reduce the pool size of available wiki editors to >>>>>> just those with commit rights. Given that committers have openly >>>>>> stated their reluctance/dislike/opposition to ever writing >>>>>> documentation then such solutions seem unwise and detrimental to the >>>>>> community as a whole. I'm not convinced by the logic used to justify >>>>>> these solutions that because the documentation is inline with the code >>>>>> that it's more likely to be kept up to date and accurate. I therefore >>>>>> strongly urge the community to reconsider. >>>>>> >>>>>> Thanks, >>>>>> Paul >>>>> >>>>> >>>>> >>> >>> >
