[boost] Porting Loki to Boost (next stage)

From: Andrei Alexandrescu (andrewalex_at_[hidden])
Date: 2002-04-16 17:08:38

Hello,

It seems like my message "Porting Loki to Boost (reprise)" failed to reach
its intent. In the nine kilobytes of the message there was a sentence that
was widely discussed.

Many people here have strong opinions and strong ways of venting them. I am
the same. The way I express my opinions and my vested interest in Loki sends
the wrong message, being interpreted as a "take it or leave it" attitude
with regard to porting Loki to Boost. This is not the case, as the 9K of my
other message tried to convey.

Furthermore, my criticisms addressed to other libraries might be seen by
default as a comparison of those libraries with Loki, even though Loki
doesn't even provide similar facilities, or provides them in a very
rudimentary fashion.

So I would like to make a new proposal.

How about finding a lead, a person with enough enthusiasm and motivation to
take relevant parts of Loki to boost, and have that person collect opinions
and make decisions? That would have many beneficial effects. One is that my
voice will be probably better heard, and my points better taken, if I was
just a kibitz among many others, rather than the sole (or principal) pioneer
of Loki. Another benefit would be that the lead (at best a boost vet) could
better figure out the best ways of integrating what boost has with what Loki
offers.

Again, at the risk of saying the "fatal one sentence" that will let this
thread drag forever, I believe the components of Loki that could add value
to Boost are: Smart Pointer, Visitor, Double Dispatch, Singleton, Object
Factory, Abstract Factory. There are a couple of features in Functor that
might suggest some changes in Boost's functor.

At this point it's become irrelevant to me what typelist infrastructure
would be best to use. If anything else, the long discussion that's just
cooling down now is a witness that different people have different
preferences.

So, does anyone volunteer to be a lead? The proposed agenda is: start with
Smart Pointers (a widely discussed component), then Visitor (a nice
component that can benefit many other areas of boost), the two factories,
Double Dispatch, and Singleton. The agenda is subject to change... by the
lead :o).

Also, a few people already expressed interest in participating at various
levels. I'd be glad if they would reply to this message, telling whether
they agree with the proposed approach, so we can build a list. Thank you.

Andrei