Towards a Modern Communications API
Michael Demmer (UCB), Kevin Fall (Intel Research), Teemu Koponen (HIIT
/ ICSI), and Scott Shenker (ICSI / UCB)
The key distribution is an important part but it
is not here. It is outside the scope of this work.
In the web page example, how do you handle
dynamics? Isn’t it this PUT/GET/PUT/GET as it is done now?
Yes it is, but players would not use the same
approach if we have new API. It would
change the way we publish.
Dave: it looks like we take out the notion where
the data is stored. It is not clear who owns the data with new API.
We thought about custodial services that would
oversee where the data is stored.
Some aspects might be expressed themselves and
some aspects in the implementation.
When you do a get on the messages
that are stored somewhere and when there is no ownership on the data,
then you could have DoS on such system bu creating a lot of messages
with huge lifetime.
It is not a prerogative of application but of an
John: do you really think it is
ideologicaly pure? don’t we mash a few ideas? VanJacobson tried to do
the same. Do you have some room for improvements? The essenece is here
on data driven model and versioning mixes everything.