[nanomsg] Re: New release?
- From: Drew Crawford <drew@xxxxxxxxxxxxxxxxxx>
- To: nanomsg@xxxxxxxxxxxxx
- Date: Mon, 27 Oct 2014 21:26:14 -0500
For what it’s worth, I had exactly the same reaction. In fact, msgpack has
recently fixed this misfeature in the specification, although that fix has not
actually trickled down to the public implementations as of yet.
I have an implementation of my own that is one of the only ones that follows
the new spec and correctly differentiates between strings and bytes.
However the fact that I had to do that, is a testament to how bad the
packing/serialization ecosystem really is.
Drew
> On Oct 27, 2014, at 9:13 PM, Matthew Hall <mhall@xxxxxxxxxxxxxxx> wrote:
>
> On Mon, Oct 27, 2014 at 04:49:46PM -0700, Garrett D'Amore wrote:
>> In fact, nanocat & macat both do msgpack serialization, IIRC.
>>
>> These details definitely belong in the app layer above, not in nanomsg
>> itself, IMO. Don’t tie the protocol to a particular packing format, but
>> keep it agnostic, because whatever you try to choose, it will be wrong (for
>> some applications).
>>
>> - Garrett
>
> FWIW, I thought msgpack was awesome at first. Until I discovered that, unlike
> YAML, it didn't by default support a notion of a textual string and a binary
> string inside of the protocol, though many other systems like YAML and ASN.1
> easily allow this. After I found that out, my brain had a huge core dump
> (ulimit -c 0) and I abandoned msgpack until they fix this misfeature.
>
> Thus I fully support the consensus previously reached in this thread, that
> having hard dependencies on any of these serialization systems is a Bad Thing
> (TM) since most of them are buggy.
>
> Matthew.
>
Other related posts:
