在 2006/7/12 上午 12:57 時,chromatic 寫到:
On Tuesday 11 July 2006 21:45, Audrey Tang wrote:If you think PIR is a language for people to write manually tocode applications in, _and_ it has some legitimate use for deleting filesin :immediate blocks, then your argument may make some sense.Come on, Audrey! That's a strawman argument. The point is emphatically NOTthat people should be able to delete files in :immediate blocks. As I said in my previous message, the same problem potentially existswith :init -- and saying "OH NO THE PARROT DESIGNERS WANT TO REMOVE YOUR HARD DRIVE!! THEY ARE EVIL AND WRONG (and they break static analysis)" is notproductive.
No. With :init, the step of .pir->.pbc, as well as all tools that works with .pir without running it -- including the very important "parrot -o" -- are safe.
The :init is part of the run time. We already know that the runtime needs a sandbox at some point. What we don't know is that why the PIR compiler -- supposedly something that's run only when HLL compilation is done -- would need a special feature for hand-written code alone, and very error- prone at that.
I can see use cases for :init to interact with the user's environment; I cannot see use cases for :immediate to interact with who-knows-what's environment, and break the otherwise intact "same .pir compiles to the same .pbc on the same
environment" property.Allison and Chip already elected to wait a while and see if someone finds a use for it, that actually requires unbounded evaluation. I now agree with them.
Please stop it.
Er, that's what I'm planning to do. Thanks, Audrey
PGP.sig
Description: This is a digitally signed message part