[zeromq-dev] what happens to the bindings with libzmq libzapi ...

[Pieter Hintjens]

On Tue, Apr 26, 2011 at 1:00 PM, Martin Lucina <mato at kotelna.sk> wrote:

> However, I do not understand your constant de-emphasising of the core C
> APIs and your positioning of libzapi as the official high-level C API.

First off, these are not my decisions. Everything I've listed here was
discussed first on IRC, and only after Martin's approval have I done
anything.

Your opinion is useful but it's not the first time you come to a
thread rather late, and then complain about decisions already made and
action already taken.

> Normally, a product such as libzapi would go through phases just like e.g.
> the language bindings. You announce it, people use it and contribute, and
> *if and only if* a need is established by the community that an official
> "high-level C API" is required, then libzapi would be established as such.

Uhm, no. No other language binding has to argue for legitimacy except
by being used. As libzapi is used, and exists, it's as legitimate as
it can be, until people want to fork it and/or make competing layers.
Which they are of course welcome to do.

> I do not see that happening here and I agree with the sentiments in this
> thread that the proliferation of APIs is a bad thing and will cause us harm
> and confusion in the future.

If you have a better answer to the actual problems that this
proliferation is addressing, please bring them forwards.

In case you forget what those problems are, they are principally:

* Martin has the ambition of removing functionality that people are
actually using from the core API.
* The lack of a formal high-level C API forced the creation of a
complex undocumented API for C apps, as part of the Guide. I'm
actually using C and need a HL API, and thus have made one.
* Other layers such as the C++ API, and the devices layers have been
throttled and basically left to die as part of the core API.

If you are simply going to dislike my work without constructive
contribution, forgive me if I ignore further comments from you.

> I would like to request discussion on this and would prefer that zapi is
> not positioned as "the official high-level C API for ZeroMQ" until/when/if the
> community agrees that this is a good thing.

You have some reason for raising the barrier higher for a C API than
for other languages?

Where are these new requirements documented? What other API has passed
this new set of requirements?

-Pieter