[REVIEW] Ch4. Branching and Merging - Advanced Merging

[Paul Burba]

On Thu, Jun 12, 2008 at 1:39 PM, Ben Collins-Sussman
<sussman at red-bean.com> wrote:
> Having read your article a month ago, Paul... I have to say, it made
> my head spin.  I know svn inside and out, though not the
> merge-tracking stuff -- and the amount of detail and edge-cases was
> bewildering.  I mean, the stuff is *interesting* and cool, but is this
> stuff an ordinary user is going to need to know?  Much of it felt like
> implementation details.  That's why I deliberately left the svnbook so
> simple.

Hi Ben,

If you've read that article* and made the call to exclude those topics
from the book than that's fine.  I only listed those things to make
sure nothing was inadvertently overlooked -- Sorry to put so little
faith in you all!

* Honestly I wasn't sure anyone had read read that beast and don't
blame them if skipped it.

> It's a hard call to make, really.  One of the goals of merge-tracking
> is to make branching and merging "automatically do the right thing".
> svn 1.5 is a first step in that direction, and we all know that
> there's a whole lot of ground to cover.  But my understanding is that
> as we begin to cover that ground, users *shouldn't* have to learn
> about inheritable vs. noninhertiable mergeinfo, or how mergeinfo is
> ellided, or what 'implicit' mergeinfo is -- heck, they shouldn't have
> to look at svn:mergeinfo at all!   So my worry is that chapter 4 is
> already overwhelming in terms of content, and adding this stuff would
> be (1) way TMI

Ok, hard to argue that.  It is certainly true that trying to talk
about these topics in a brief but meaningful way is almost impossible
(for me at least).

> and (2) eventually become non-useful as merge-tracking
> improves.
>
> The tradeoff, of course, is that when users try to do something in 1.5
> and the merge-tracking feature bites them -- they have no
> documentation to turn to.  Well, not really -- they have your paper,
> and the users@ list.  But the book won't help them untangle themselves
> from edge-cases.  On the other hand, I somehow doubt that if all the
> implementation details of svn:mergeinfo *were* in the book, that
> they'd be able to untangle themselves either.

That's actually quite a compelling argument for skipping this stuff.
I convinced myself that understanding this stuff is useful many users,
but that is probably my own bias for being so close to it for so long,
or less flatteringly: "The stuff I've been working on is *really*
important and everyone should feel my pain!" :-)

> (Well, maybe only a
> teeny percentage of brainiac users might...)  So on the whole, I'm
> thinking it's better not to overwhelm the vast majority of users with
> details that aren't likely to help them, only intimidate them further.
>
> Perhaps the best compromise for now is to put a sidebar in chapter 4
> which alludes to the mega-complexity of svn:mergeinfo,

Be sure to include a skull and crossbones graphic.

Paul

> and points
> people to the collab.net paper?
>
> On Thu, Jun 12, 2008 at 12:16 PM, Blair Zajac <blair at orcaware.com> wrote:
>> Paul Burba wrote:
>>>
>>> Hi All,
>>>
>>> Took a look at the advanced merging section of chapter 4.  Everything
>>> that's there looks good, my only concerns are what is not not
>>> mentioned:
>>>
>>> 1) The fact the svn:mergeinfo is inheritable (particularly the
>>> limitations of this in mixed-revision working copies).
>>>
>>> 2) Non-inheritable mergeinfo
>>>
>>> 3) Empty mergeinfo
>>>
>>> 4) Copy/move can produce svn:mergeinfo
>>>
>>> 5) Implicit mergeinfo (i.e. "natural history")
>>>
>>> 6) Mergeinfo elision
>>>
>>> Is it worth mentioning of these subjects or do these get into too much
>>> detail for the book?
>>
>> Not being an author on the book, I don't have much say, but I would like to
>> see these topics discussed.  Otherwise we'll have to forward people to read
>> your article :)
>>
>> http://blogs.open.collab.net/svn/2008/05/subversion-15-1.html
>>
>> Blair