[Twisted-Python] Task-based documentation started

[Tom Davis]

On Tue, Feb 1, 2011 at 5:15 PM, Glyph Lefkowitz <glyph at twistedmatrix.com>wrote:

>
> On Feb 1, 2011, at 10:53 AM, Tom Davis wrote:
>
> On Tue, Feb 1, 2011 at 12:58 PM, Glyph Lefkowitz <glyph at twistedmatrix.com>wrote:
>
>>
>> On Jan 31, 2011, at 11:41 AM, Tom Davis wrote:
>>
>> Thoughts?
>>
>>
>> Priority #1 for most people who are enthusiastic about documentation is to
>> come in and write a ton of additional documentation.  But this is a lot like
>> trying to fix a large, broken, untested system by writing a pile of new,
>> untested code, because "this time we'll get the design right".  What were
>> the problems with the way the previous documentation got written?  How did
>> we end up with this mess, and what is going to be done differently this
>> time?  Most importantly, *what is the metric by which we will judge this
>> new documentation to be better?*
>>
>
> I'm still really interested in concise answer to that last question there.
>  What are the priorities for what you're trying to improve?  What you've
> listed here is "the steps you're going to take in order to improve it" but I
> have yet to see a lucid description of the problem in detail.
>

I think the problem is largely one of organization (as it relates to
"accessibility", e.g. how / under what context do I get linked to such and
such thing?) and outdated documentation. I think the individual problems are
all very well-known and while there is not some overarching problem with all
the documentation that requires starting from scratch, reorganizing and
restructuring what's there may make it look like I'm trying to do that.


>
> Those tutorials aren't new at all; they were taken (often verbatim) from
> the "using twisted.web" document. Writing a ton of new documentation really
> isn't my goal at all. My goals are threefold:
>
>    1. Reorganize existing documentation in a way that makes it more
>    accessible.
>
> Accessible to whom?  How will reorganizing it make it more accessible?  One
> way to interpret this statement is that your goal is to remove all diagrams,
> because the documentation is currently not accessible to the blind.  I'm
> pretty sure *that's* not what you mean, but it's a good example of how
> ambiguous this statement is :).  (Plus, the couple of blind developers I've
> talked to about Twisted didn't seem to have a problem with the docs...)
>
>
>    1. Edit existing documentation to conform to a task
>    (howto/tutorial/etc.) vs. expanded learning model. (the whole instant
>    gratification thing and all that)
>
> OK, I'm on board with that.  Except that in order to understand the
> tutorial documentation you need a good backing of reference docs, so it's
> not like you can just choose one over the other.  And we have lots of
> tutorial docs (c.f. the infamous finger tutorial) which are in the tutorial
> / task-oriented paradigm but don't teach much that's directly useful.
>

Finger is a really bad example of what a howto should be; it doesn't teach
you how to achieve some specific, common goal quickly and easily. The names
howto <http://twistedmatrix.com/documents/current/names/howto/names.html>is
a good example of a howto that tells me exactly how to do something in as
few steps as humanly possible. It could use a reference to wherever all the
functions (SOA, A, NS, etc.) are coming from and links to some follow-up
documentation of the *names* project in general, but I digress.

>
>    1. Gradually update / replace code listings with "current best
>    practices" examples that are tested.
>
> No arguing with that at all, that sounds great.
>
> I know you want an actual *layout* for the reorganization.
>
>
> Not really.
>
> What I want is a clear understanding of what you intend to change.  I
> usually get that by reading a diff: it's easy to read a diff and see what
> changed in the old versus the new version of something.  I can tell if it's
> an improvement without having to digest the entire content.
>
> But the structure of <http://docs.recursivedream.com/twisted/> is a
> sprawling mess of placeholders and half-finished ideas.  It's different from
> the existing documentation in lots of ways; it has a completely different
> stylesheet (which I assume is some standard Sphinx thing we will get rid of
> with the consistent theming work in Kevin's sphinx transition).  I don't
> know what I'm looking at or how to appreciate it.
>
> This is why I started off by complaining about the separate git repository.
>  If you're going to work in a separate repository and a separate format and
> not produce diffs that I can skim, then it's *very* hard to comment
> intelligently on your strategy, and you need to take an immense amount of
> care to call out the sections you consider complete, what exactly you want
> feedback on, and areas where you have or haven't added new content (so you
> can't be blamed for issues in old content that you're not trying to update).
>
> Note that diffs against Kevin's sphinx output would basically satisfy this
> same requirement, even if that output isn't really complete and those diffs
> would need to be re-created when the final conversion occurred.  At least
> there would be a set of deltas to look at rather than a whole new structure
> which is mostly scaffolding.
>

I'm just going to start making diffs against Kevin's lore2sphinx output as
suggested. He's given me instructions on how to generate the docs and I'll
attempt to create specific patches and tickets against that output to the
extent possible. My hope is that this method will allow the project to get
somewhere that doesn't result in me ending up with a bunch of third-party
Twisted documentation because nobody could ever really agree on why it was
an improvement in the first place.

And on that note, perhaps it won't be an objective improvement. I'm open to
the possibility that the documentation is sufficient for most people and I'm
just an exception to that—or that my "improvements" are only really helpful
to me. But I'm certainly going to keep looking for ways to suitably prove
the general case, for everybody's benefit; I have no desire to end up with
something that only benefits me.


>
> I want to give that to you, but it requires going through all the
> documentation and making sure it fits in the reorganized layout or making
> sections for it when it doesn't which is far more time consuming than
> providing a layout for the "task" piece, as I did here. I'm basically just
> saying "How-tos should have a standard layout with these sections; here's a
> stub of one such how-to as incompletely adapted from the current
> documentation".
>
>
> Even going back and reading your original message to this thread now, I
> can't tell that this is what you were asking for review of.  Even now I'm
> not entirely sure what you mean.  Do you mean that you want feedback on the
> sections on <
> http://docs.recursivedream.com/twisted/projects/web/tasks/serve/index.html>?
>  Or the way that <
> http://docs.recursivedream.com/twisted/projects/web/tasks/serve/serve.html>
> is split out into separate tasks?
>
> If you are interested in getting a review of the template or the outline,
> please write up the template itself with a description of what should go
> into each section.  Then you could link from the template to
> web/tasks/serve/index and say "and here's an example of some existing
> documentation applied to this outline".  But since I have no idea what the
> generic purpose of each section is, I can't comment on whether the sections
> are a bad idea or the classifications of certain sections of existing docs
> is a bad idea or if it's just too incomplete for me to comment on.
>
> Plus, such a template would serve as a critical tool for new documentation
> authors (of which I hope we get a few), allowing for a consistent style to
> be followed by multiple authors writing task-oriented documentation for
> different parts of Twisted.  The lack of such meta-documentation is the root
> of many issues with the current docs: when we're writing code for Twisted,
> we have a very clear idea of what the coding standards are, but when we
> approach documentation, the individual author just writes whatever random
> style happens to suit them on that day.  There's no guidance.
>
> I think most of this is a huge misunderstanding as to my goals both in
> general and for the particular section critiqued here. Perhaps I should have
> completed it before putting it up for review. I just felt it was better to
> get *something* out here just in case everybody hated the entire direction
> rather than spend a ton of time slicing the current web server docs the way
> I personally feel they should fit together in order to properly serve the
> audience.
>
>
> Soliciting feedback is good.  I still think that these discussions have
> been very productive.  The issue I have isn't that it's unfinished, it's
> that there are no guard-rails on the unfinished sections, so it's easy to
> careen off into the stuff that I'm not really supposed to be paying
> attention to. A substantially larger document with that same problem would
> have been a disaster to try to review, so early feedback is better;
> hopefully as you produce something bigger you'll also alleviate the
> difficulty of review somehow.
>
> If all that comes out of your efforts is a new, different pile of
>> documentation, that's not *entirely* without value; different audiences
>> understand things in different ways, so someone might pick it up and
>> understand Twisted as a result.  But I very much doubt that's going to move
>> us from a general impression of "bad docs" to a general impression of "good
>> docs".
>>
>> More specific criticism:
>>
>>
>>    - "This is a Twisted Task"?  I feel like I'm about to start reading
>>    about a Task object or something related to the twisted.internet.task
>>    module.  Or maybe that this is an exercise.  A document explaining how to do
>>    a task is rarely called a "task" itself.  Mostly they're called "How-To"s,
>>    actually.  I understand you're trying to get away from old terminology but
>>    this seems awkward and forced.
>>
>> I just started using this terminology because it seemed common within our
> conversations. How-to/tutorial would likely be less confusing.
>
>
> OK.  For what it's worth I share your desire for a new / better term but I
> can't think of one.
>
>
>>    - "You should be familiar with Resources"?  That is really broad and
>>    wide-ranging and should be a hyperlink to another tutorial.
>>
>> It was just a placeholder; it should link to a section in the web server
> overview docs that talks about Resources. There are other things on that
> list, too (Site object, yada yada). Although, like I said, it's merely
> incomplete.
>
>
> I suppose my comments above about meta-documentation and a deeper
> explanation of the template are my response to this; if it were clearer what
> the structure were really supposed to be representing it would be easier to
> ignore minor flaws like this or recommend improvements
>
>
>>    -  Plus, I think that most people interested in these tasks will want
>>    to learn about how to get simple tasks done first, before moving on to a
>>    more abstract / theoretical understanding of the model behind them.
>>
>> I completely agree. That's my whole point with starting off using
> *twistd*, despite its limitations or whatever.
>
>
> Obviously from my other message I'm a fan of this approach.
>
>
>>    - More importantly: this is a very wishy-washy definition of the
>>    audience.  Is it for system administrators?  Software developers?  DevOps
>>    people?  Graphic designers?  Power users?  What level of experience do they
>>    need with Python?  With networking?  With HTTP?
>>
>> The latter should all be under "Prerequisites" whereas "audience" should
> probably be a new howto standard section that's currently overlooked
> (perhaps above "Prerequisites"?)
>
>
> It seems to me that "audience" is the section and "prerequisites" should be
> a sub-section of that.  After all prerequisites are really attributes of the
> audience, not the document :).
>
>
>>    - "Other configuration options are available"?  This should be a list
>>    of links to other documents that might help with some of those options.
>>     Documentation which says "and then you could do something else" without
>>    telling you what else or why is needlessly confusing.
>>
>> I agree. There are lots and lots of lines to connect even from such a
> simplistic tutorial. I wanted to get feedback on the *layout* rather than
> the final web project documentation which should certainly include links to
> twistd option docs and probably a half dozen other things I haven't linked
> to yet.
>
>
> (Again, go write up the layout explicitly first, and then I can provide
> more useful feedback!)
>
>
>>    - The WSGI example totally glosses over how you get your code onto
>>    sys.path (PYTHONPATH etc), says that 'helloworld' is a module/package (which
>>    is it?  why are both options given?).  This is an *extremely* confusing
>>    area for newbies, which is why it actually makes a bit of sense to me that
>>    the web-in-60-seconds WSGI tutorial starts with a callable defined in the
>>    .tac.  Not a good practice in the long term, but it allows the user to
>>    immediately get some feedback and have some notion of how things are wired
>>    together without having to debug Python's import system first.
>>
>> Fair enough. This point was also not covered in the official docs, where I
> got the instructions from. Regardless, I have no intention of including
> how-tos that gloss over *anything; *either they work "out of the box" or
> are edited until they do.
>
>
> I guess this is another nitpick that wasn't really about the layout.
>  Sounds like we feel the same way, at any rate.
>
>
>>    - .rpy scripts are for deployment, not debugging; I was a little
>>    concerned seeing that bullet point on the outline.  (See <
>>    http://twistedmatrix.com/documents/10.2.0/web/howto/web-in-60/rpy-scripts.html
>>    >.)
>>
>> Huh? The docs you link to claim:
>
> So, while rpy scripts may be useful for testing out ideas, they're not
>> recommend for much more than that.
>
>
> Am I missing something here?
>
>
> Well, "testing out" and "debugging" aren't quite the same thing.  And I'd
> probably recommend RPYs for more than that :).  I was linking to that
> document just to provide an expanded explanation of the functionality;
> personally I find it obvious why RPYs are super useful from that example, so
> I mostly just ignore that comment.  (Perhaps this doc needs to be updated,
> but let's please not hijack this already-sprawling thread for discussion of
> the merits and flaws of RPYs...)
>
> My hope is that this reply clears up some confusion. However, feel free to
> let me know if you think I simply don't have the mindset necessary to
> "improve" the documentation in a way that is desirable. I had a hard time
> learning Twisted in the beginning and to this day some really helpful
> abstractions and just available projects / solved problems elude me.
>
>
> This sounds like I'm discouraging you.  If that's the case, feel free to
> just tell me to sit this one out and wait until you're further along in the
> project.  I am trying to provide feedback to be useful to your efforts, but
> if you don't find it useful, then you can carry on with feedback from people
> who really need the docs.
>
> I think the reason for this is that I think a lot like Jacob <
> http://jacobian.org/writing/great-documentation/> when it comes to
> documentation and that's the sort of stuff I need to have in order to truly
> grok something. Maybe that's what we all want here. Maybe it isn't, though.
> And I'd really love to figure out which it is before I drive forward on
> this. If we really are on the same page (and I'd certainly like to be),
> we've got some hellish communication issues ;)
>
>
> I'm a big fan of Jacob's in general and that essay series in particular as
> far as it comes to documentation, so I don't think that's the issue.
>  Obviously his methods have yielded a result that is well-regarded in the
> Django community.  And I think we are on the same page about many things,
> and we don't *need* to agree about everything.
>
> I'm still quite excited to have you working on our documentation.
>
> (Except for his comment about API documentation generators; I think that
> epydoc/pydoc is hugely useful.  But I do agree that one shouldn't just run
> an auto-generator to spit out a rehash of the source code; one needs to
> write a ton of documentation in an appropriate format to be consumed by such
> a tool in order for it to be useful.)
>
> Thanks and good luck,
>
> -glyph
>
>
> _______________________________________________
> Twisted-Python mailing list
> Twisted-Python at twistedmatrix.com
> http://twistedmatrix.com/cgi-bin/mailman/listinfo/twisted-python
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://twistedmatrix.com/pipermail/twisted-python/attachments/20110206/37d9e313/attachment-0001.htm