with webcomponents and isolated dom window objects
we can do one component can 2
other can 3 and so on
but you need really to think a lot because the biggest problems are globals
but to make it short yes its today possible to run diffrent steal + canjs versions
and even mix that up
but its more complex then it should be
thats not a problem
steal assigns to window as global
so you simply need 2 diffrent window objects
for example a iframe generates a secund window object
and there are also diffrent other methods to get that
and a iframed window can controll outer window
iframe (where is the puke emoticon) :P
I thought iframes are dead... they certainly are hacky though
but more for isolation and controll flow
not for displaying view able content in them
i did create a secund html diffrent steal canjs version
that gets imported via html link rel import
and then this window syncs up with the importing document
so your index html requires legay.html
and legacy.html attaches to your index.html view
legacy sorry not legay
lol
and it does exactly that
but not in its own window
in this POC
but there is no need for that
import and iframe are good methods
this way every thing works as befor
thats it
I was thinking about this issue, that I've asked before and got the idea... what if I list can as peerDependency for my components ? I mean I know that my component should work with ...can 3.x and the only thing that might brake my component is some API change which is fairly rare case. But if I do this way I'll avoid this "2 versions of something found" because entity-details-component will use "host" packages. Well at least I expect it to do this :)
@phillipskevin et al. In an attempt to learn about bundlers other than steal with canjs, I put together a small demo @ https://github.com/DaveO-Home/embedded-acceptance-tests. I think it works....
So I see that react is more performant in this example now? https://output.jsbin.com/wotevub/2 Are you guys seeing the same?
