On Sep 17, 2008, at 6:16 AM, Malcolm J. Currie wrote: > I'd be reluctant to change the API of such a low-level library that's > widely used. While we could change the Starlink code, it may have > unforeseen impacts elsewhere. Likewise the different token operators > have been around for twenty-odd years, and I'd be wary of changing > those too. Well, the aim was to *not* change the API. But it's simply too much effort to try to work out whether sprintf will be required. I went into this assuming that variable args could be counted but since it's not possible to know how many optional arguments are being used the string has to be parsed to try to determine whether any are expected. If I could work out whether any optional args were present then it would be trivial to retain compatibility with the existing API without any hoops (simply saying that optional args can only be used if you are prepared to escape % and not use %ET etc). -- Tim Jenness Joint Astronomy Centre