On 12 Jan 2017, at 2:13pm, Donald Griggs <dfgri...@gmail.com> wrote:
> Further, Dr. Hipp and his team won't have to deal with howls of "it's > broken" when such a version is released. Just taking the Devil’s Advocate position here, a lot of programmers would argue that inconsistent operation is broken. They can deal with any consistent behaviour for something they didn’t specify, but not with a program which does one thing one day and another another day. I remember having a related problem with a mainframe database engine years ago. This multi-user system had a client/server design and all data operations were performed by a program running centrally. The server created various temporary indexes as it needed them and it kept them in memory until it needed the memory for something else. If it wanted to read every row of a table, and an index for that sort was already in the cache, it would use it. But if no index was already cached it had to make a new one up, which could take anything up to a couple of minutes. So your program could perform differently depending on which sorts or searches /another/ user had done recently. And you had no way of knowing what that was. This lead to complaints since creating an index could take up a lot of processing time and input/output, and those were charged to the department which ran the program. A programmer had no way of knowing how much running his program would cost the department and department managers hated this. Nevertheless we eventually settled on the same answer you’ve see in this thread: If you want something specific, specify it ! Maybe you’ll get lucky and get a 'free' index. Maybe you won’t. Simon. _______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
