From: Joe E. <jen...@fl...> - 2006-11-03 05:28:11
|
Steve Landers wrote: > Joe English wrote: > > This would be an excellent candidate for tcllib, but I don't > > believe it belongs in the core. > > [...] > > Good points Joe, but I don't see this as an "either/or" situation. > > I'm more than happy for platform to go into tcllib and evolve > quicker, but strong believe that it needs to go into the core as > well, so developers can rely on it being present. Putting this -- or anything, for that matter -- "into the core" doesn't help all that much when it comes to universal availability, since it leaves a large class of end users behind. Namely: Unix users who rely on whatever Tcl build comes with the OS. There's a big delay between when a facility goes into the core and when *all* developers can rely on it being present. Most of us have to wait for: + The next Tcl release -- around 6-8 months for dot-dot releases (and upwards of four years for dot releases, but that's a different problem). + The OS distributor to pick it up -- a couple of weeks if they're paying attention, longer if they're not, or if they're in their own freeze phase. + The OS distributor to make a new release -- typically 6-12 months, much longer for "enterprise editions" and for Debian. + The end users to upgrade their OS -- this can take anywhere from 24 hours (Gentoo users and other bleeding-edgers) to "never" (people working for companies with obstructive QA policies). Now if the platform package were the kind of thing we could get right the first time, then sure, put it in the core and _eventually_ everyone could rely on it (some sooner, some later). But it's not: by its very nature, [platform::identifier] is going to be perpetually incomplete and frequently out of date. No matter how extensive the implementation, it's going to be obsolete the moment somebody ports Tcl to a new platform. Of course putting it in tcllib won't make it universally available either, but it *will* make it easier for people who need the latest version to get it when they need it. > Remember that the TIP authors are at the pointy-end of delivery multi- > platform distributions and extensions - and only too aware of the > problems that entails. Y'know, I've done a bit of development and deployment myself. --Joe English jen...@fl... |