On 2/7/11 4:43 PM, Scott Wilson wrote:
On 7 Feb 2011, at 14:22, Marcos Caceres wrote:
On Mon, Feb 7, 2011 at 1:46 PM, Scott Wilson
<[email protected]> wrote:
I really like the Kill Switch/EOL idea and having a "type"
attribute to specify it, but I'm concerned that the Patch type
could be a bit more problematic to get consistently implemented.
Understood. What concerns are you having or what interop issues do
you foresee?
Principally the handling of the various update states, rollbacks
after failing to apply patches, problems with
multiple-version-spanning patch updates that kind of thing.
Also when we unpack a widget and ready it, its no longer exactly the
same as the input .wgt so we'd have to apply the patch against the
originally imported package rather than the actual installed instance
and then load it again or the patch won't take - so we may as well
update the whole package anyway.
Both excellent issues.
Its not a bad idea in principle, but potentially a lot of code to
save a few kb of downloading.
I agree. For small widgets this is not an issue. It's for big widgets
where it becomes a problem. It might be that the user agent could do
negotiation (e.g., "I don't support patches and have plenty of
bandwidth, just send me the whole thing").
--
Marcos Caceres
Opera Software
