[racket] Poll: Does anybody besides Doug use 'plot'?

From: Robby Findler (robby at eecs.northwestern.edu)
Date: Sat Oct 1 18:10:08 EDT 2011

On Sat, Oct 1, 2011 at 4:28 PM, Eli Barzilay <eli at barzilay.org> wrote:
> Two hours ago, Neil Toronto wrote:
>> On 10/01/2011 12:18 PM, Neil Van Dyke wrote:
>> > Option #1 seems like an easy way to go, from what I can see.
>> Okay, you've almost got me convinced again. I'm so
>> wishy-washy. Sorry, Eli.
> Just to make my point here (since I avoided this thread): the
> important question from my POV is of work that you will need to deal
> with, vs benefit from that work.  This poll thread, as I suspected,
> shows that there are no important uses of `plot' from people who would
> find it difficult to replace (require plot) to (require plot/compat).
> (And that means that even Robby is not such a person.)
> Now, maintaining a compatibility collection named `plot' means that
> there will be more work dealing with it, and as small as this might
> seem, the expected benefits of that are so far zero.  But it's still
> your choice, as the person who needs to do all this work.  As far as
> the rest of us go, we're losing some potential improvements that you
> could have implemented in that time.

Why is this work lessed by changing the name from plot to plot/compat?

> And to go up a level -- all of that is not any point that can be
> generalized to anything else.  I consider these cases as each
> requiring its own decision.  It only happens that in this particular
> case the benefits are so small, that any work seems wasted.
> Especially given the history of the collection as undermaintained,
> hard to use, and problematic to build.  This is *completely* different
> from Matthew changing something in the core language, where making
> both the new and the old APIs work together -- the changes there can
> be much more subtle (since they're almost always gradual, not a new
> reimplementation), breaking stuff there is harder to follow (since
> it's often not just changing a `require'), and even finding places
> that would need to be adjusted is hard (which code will break if
> `define-struct' gets the same semantics of `struct'?) compared to just
> looking for (require plot).
> At this point, the amount of noise generated by this discussion is
> definitely too much.  The reason I suggested a *poll* was to have a
> clean measure of the benefits/cost amounts, not to start a discussion
> on the philosophical aspects of software changes and how to deal with
> them.  It's not the first time that such changes happened -- and
> frankly, I'm surprised at how much noise this thing has lead to.
> Compare that to the web-server change wrt responses -- this is a much
> more established library which is used by much more real code, yet
> adding a `compat' module went almost without a peep.

For the record, I complained about that. (And then later was bitten by it :)


Posted on the users mailing list.