On 14-Jan-02, Gregg Irwin wrote: > << But isn't this incredibly slow compared to running the function > locally? >>> > > Yes, but...it depends on your connection, the overhead of the > marshalling, and the amount of time spent in the function, as well as > the frequency of the call. > > Even making an out-of-process call on the same machine is much slower > than an in-process call, and you *don't* want to call simple built-in > routines, in a tight-loop, through a slower pipeline. However, if > you're analyzing weather data for a national laboratory and you can > call a routine that runs on their Cray and just gives you back a > result, that's what you should do. > > IOW YMMV. :)
I agree. But there is a tendency currently to pretend that everywhere on a network is just the same as a location in your own computer. There are always going to be delays which are vastly greater than the delay in accessing a local drive. And a WAN is of course slower to respond than a LAN. So remote calls should only be used for routines which take more than, say, 15 seconds to run, such as your weather example. You can argue over the 15 seconds figure if you like. This is not meant to be a hostile criticism of Rugby, just to point out that there is no magic. Regards -- Don Cox [EMAIL PROTECTED] -- To unsubscribe from this list, please send an email to [EMAIL PROTECTED] with "unsubscribe" in the subject, without the quotes.
