From: David Abrahams (dave_at_[hidden])
Date: 2004-07-18 02:18:48
"David B. Held" <dheld_at_[hidden]> writes:
> David Abrahams wrote:
>>>Ah, but you yourself have said that the guarantees do not form a
>>>hierarchy, so there is no proper notion of "stronger" with respect
> >>to them. ;)
>> Whaaaa??? I never said that! Or I was delusional when I did.
> Well, perhaps you remember a thread in which I suggested another
> guarantee that I called the "smart guarantee" which was a cross
> between the strong and the basic guarantee. You argued that one
> could invent any number of guarantees that offer different promises
> and which could not be ordered in any purely objective way. It was
> a fairly convincing argument at the time. Are you going to retract
> it? ;>
Not at all. But I never said the marked thing above. Certainly any
guarantee that only adds requirements to another is stronger. It's a
partial order. There are plenty of pairs of guarantees that have an
ordering relationship, and plenty of others that don't.
>>>Saying that it maintains its invariants instead of saying that it
>>>gives the basic guarantee does not convey the right message, in my
>>>opinion, because many programmers obviously do not feel a need to
>>>maintain invariants in the presence of exceptions (or they would write
>>>more exception-safe code).
>> I don't see the connection. What message do you think it should
>> that it does not send?
> That it maintains invariants *in the presence of exceptions*.
Quoting my original suggestion:
>> The library maintains its invariants and does not leak resources
>> in the face of exceptions. Some library operations give stronger
Perhaps we have an "and" associativity/ambiguity problem here?
It meant the same as this when I wrote it: ;-)
In the face of exceptions, the library maintains its invariants and
does not leak resources.
> The latter part is implicit, but if the user isn't thinking in terms
> of exception safety to begin with, then they may well miss that
> part. When you say "provides the basic guarantee", if the user is
> not familiar with exception safety, she will at least have an
> opportunity to investigate it. Of course, you could just say both.
Which is what I suggested in another message.
-- Dave Abrahams Boost Consulting http://www.boost-consulting.com
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk