Normand,<br><br><div class="gmail_quote"><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"> Olivier: I don't think idgen = ReplaceDuplicate means delete the existing<br>
feature in the database that has the same id as the one in the insert<br>
request. Of my understanding of the OGC 04-094 table 3 of the<br>
specification document, it means more to replace the id of the request<br>
with a new generated id.</blockquote><div><br>Yeap i really think you're right !<br>Need to be fix in SVN<br> </div><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
Next, I don't think we need to generate a random id when idgen =<br>
Generatenew or when idgen = ReplaceDuplicate. For my part I would take<br>
the id of the last feature in the layer and increment it by one. I think<br>
it will be simpler to manage.</blockquote><div><br>Yes for serial PostgreSQL type it could be done quite easily, <br>but what about other data types ? :<br><br> - varchar one<br><br> - (long) int one <br> - > with collision risk on several simultaneous requests <br>
<br><br>On the other hand a reliable random id generator ANSI compliant<br>could be a pain to implement, i took any suggestion ! :)<br><br><br>--<br>Olivier<br></div></div>