Re: Problem with aborting entire transactions on error

On Dec 11, 2012, at 6:26, Zbigniew <zbigniew2011(at)gmail(dot)com> wrote:

> 2012/12/10, Kevin Grittner <kgrittn(at)mail(dot)com>:
>
>> Could an option like this be added to the server? Probably. Has
>> anyone worked on such a feature? Not that I know of. Would such a
>> feature be accepted if written? I think a better case would need to
>> be made for its value compared to the added code complexity and
>> maintenance burden than has been made so far. Since this is
>> primarily of interest for bulk loads and interactive work, and
>> there are already solutions for both of those, it would probably be
>> seen as not worth its weight.
>
> Thanks for detailed explanation. From what I see, it's rather not
> going to happen.
>
>> It's only by careful gatekeeping that the source code has been kept
>> manageable. A quick count shows that there are 1,167,732 lines of
>> code, counting only *.c and *.h files. Every added line has a
>> maintenance cost.
>
> I agree - but, on the other hand, every added line of code in the
> software of the users (programmers using PG) has its maintenance cost
> too.
> And it has to be multiplied by the number of the users, then
> again: by the number of the users of their software...
> --
>

PostgreSQL itself has a very large installed extended user base and of that group the need for your feature is likely not that large. The sub group of those that can use such a generic feature instead of coding something custom anyway is likely the larger portion as well.

While it is can be rewarding to solve a problem for the masses the ambiguity and apparent difficulty means that likely only someone paid or experiencing pain directly is going to make the effort. Ideally a large number of persons would band together and fund such an initiative (even if just for research) which would both cover development costs and give the community a better idea of the scope of the problem.

David J.