From: Karl Nelson (kenelson_at_[hidden])
Date: 2002-01-31 14:19:04
> Now I am really congused. How many format libraries do we need? Since
> I am not an expect in this area I can't make sound analisys which
> approach is better, but can't "format guys" come to some consensus
> and present one solution? Again I have nothing against the
> submittion, but I do not want to end up like Buridan's donkey.
Considering that this code was originally submitted in August 2000,
and there still has not been a consensus as to what exactly a format
should do it seems unlikely there will be "one solution".
I have strong dislikes of the reviewed format as to too
complex, too much support from stuff iostream can't handle
(and I argue it should be in iostream not the format), and its
enlisting of another operator. I also think those dislikes will
be mirrored if the reviewed library even comes before the
C++ steering committee. However, as boost
allows the author to decide when it appears to just be a
perference and I originated some of the ideas and code,
I am simply asserting influence I have neglected since my
original submission (due to now expired time constraints).
These code bases will likely converge shortly, but for now this
submission is intended to interject some new ideas into the debate
and I hope refocus the debate from "should it have format and
reorder or perhaps it should have latex like ...."
to "at what level it should support printf" and "how much performance
lose is acceptable to support printf".
This "new" code base also provides coding fodder for improving
the reviewed library and future implementations.
I would submit that format3 would be a simpler starting point for
making what some people want than the other (lets-make-a-new-
format-system-completely) code base. So it does serve a purpose.
Does boost have a policy which prevents differing implementations
of similar ideas or does it encourage them? I have asked in the
past and the answer was always resoundingly "encourage them."
Previous places where this happened resulted in only one string
class going to steering where there were many competing ideas
some of which may have been superior.
I should also note that there is one submission even previous to Aug 2000
that which didn't use printf format at all. That author as well has
tried to find support in boost which started the current thread.
Thus I suspect the debate will rage on for a while.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk