> The question is whether we want to make that docroot change persistent > across requests [1] or per-request[2]. Geoff's proposal matches the > former [1]behavior, whereas Joe's the latter[2]. > > I tend to think that per-request[2] is the sane solution. That's because > you can never know whether this value was already modified or not when > you hit a new process. So you will always try to change it, in which > case per-request[2] makes the most sense. > > the only cons is that it's different from mp1 behavior, but as the above > explains it makes no difference (I believe).
automatic per-request is definitely the way to go if we can do it - people just don't understand the unusual scoping of this field, so helping them along is probably best. --Geoff