Re: [boost] Unique identifier

From: Robert Ramey (ramey_at_[hidden])
Date: 2006-11-09 11:42:22

That's not exactly correct.  From the link sited:

"permits association of an arbitrary key with a type. Often this key would 
be the class name - but it doesn't have to be. This key is referred to as a 
GUID - Globally Unique IDentifier. Presumably it should be unique in the 
universe. Typically this GUID would be in header files and be used to match 
type accross applications."

So the extended_type_info is a USER rather than a SUPPLIER of ag GUID.

One of the main functions of BOOST_CLASS_EXPORT_GUID is to associate
at type with type with a portable GUID.  The specific GUID is left up to the
user.  The macro BOOST_CLASS_EXPORT(class name) just uses the
class name as the GUID.

So the existence of an open, portable GUID would complement
extended_type_info rather than conflict with it.

This system let me isolate serialization from having to address a number of
issues related to comming up with a GUID.  Among those are:

a) it would seem attractive to make a GUID derived from typeinfo as
provided by RTTI.  In practice this doesn't seem attractive since there
is no standardized typeinfo format.

b) namespaces:  the same type might appear in different namespaces in
different programs.  In fact, one of the uses of namespaces is to "wrap"
libraries so that it can be guarenteed not to accidently conflict.

c) Having the GUID look like a class name is attractive - but then
it runs into character set issues.

d) Having the GUID be derived from the class declaration might
seem attractive, but what bout when the class changes - we
don't want to obsolete all the programs which depend upon it -
or do we?

and others.

So what starts out seeming like a simple idea, starts to get pretty
complicated as one tries to realize it in a specific way.

So - the extended_type_info - and by implication the serialization library
takes a pass - it presumes you've got your own GUID or are happy
with using the class name.  And this shows up as a problem from time to
time as user want to use classnames including a namespace as the GUID
which conflicts with HTML.

Robert Ramey

Janek Kozicki wrote:
> Andy said:     (by the date of Thu, 9 Nov 2006 15:57:31 +0000 (UTC))
>
>>> Thanks Sebastian,
>>>
>>> The file I downloaded was GUIDv3.zip, as I couldn't find anyother
>>> (did I go to the wrong place: boost-consulting.com/vault ?).
>
>
> GUID which stands for Global Uniqie IDentifier is already in boost.
> Why include a second one? If you want to work on this, then better
> improve the existing one.
>
> For the beginning, it can be separated from lirbary
> boost::serialization (it is an extended_type_info):
>
> http://boost.org/libs/serialization/doc/extended_type_info.html
>
> In fact it is already a separate module (because it makes testing much
> simpler), so the separation will just make things more formal, and
> involve just changing namespaces.