[erlang-questions] fyi: Google protocol buffers

Christian S <>
Thu Jul 10 01:42:33 CEST 2008

> I looked at UBF a long time ago and I thought it was a very
> interesting idea. What I was wondering is why there seems to have been
> so little apparent interest in this (i.e. why hasn't it transitioned
> from the POC).

As long as we are a bit cynical in this thread, I'll share some of my cynism.

I think UBF has got no attention because the skill (in the bottom 95%
of programmers) at using parser generators and parsing is low, people
go as far as splitting text on a delimiter character (or regexp
pattern), they maybe do this further on the results, but that's it. If
more is needed, they bring in XML, because DOM parsers doesnt
intimidated the ego the same way a tool you dont know will do.

UBF, the transport format, can only really be appreciated if you know
and directly see that it is easy to implement (comparingly). If you
can see that it is very feature competent, that it has very un-muddled
data types, the plain list, the tuple, integers, strings, symbols,
8bit binary blobs (without silly base64 tricks). There is no XML
schema on top to bring you elementary data types that all
machine-to-machine transport _will_ need.

UBF, the protocol contract language, can only really be appreciated if
you get to the next level of understanding publicly exposed protocols,
implementing them, maintaining them, documenting them, designing them.
  So much about programming against an interface is "can i say this
right now?", or, "what do i need to say before what i really want?".
You dont see many interfaces targeting this kind of specification, all
you get is that they're going to throw IllegalStateException if the
programmer bothered to check premises explicitly, otherwise you will
likely get a NullPointerException and no guarantees about how valid
the internal state of the object is now.

Sorry for my rant.   ... ranting is too much fun.

More information about the erlang-questions mailing list