[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [RRG] Why delaying initial packets matters
It is indeed true that in some cases there is no such thing as a free
lunch. And keeping the Internet up and running is important. As we
invent new pieces of technology -- such as a new routing architecture in
this case -- there may effects to the other parts of the Internet system.
However, we should not jump from this to the conclusion that we do not
need to understand or measure the effects to applications because "we
have to do it anyway". We do need to understand the effects, for a
variety of reasons:
- We want to choose the best solution, so understanding the differences
between approaches on this aspect is necessary.
- If the effects are truly significant, we may have to go back to the
drawing board to invent better solutions.
- If it turns out that a certain effect is unavoidable, we want to make
sure that it is still acceptable compared to the alternative of not
having any solution at all. I.e., while fixing the Internet please don't
- If the effects can be avoided by a significant increase of complexity,
will it be worth the effort?
- Deployment incentives for people installing either the new routing
system or the needed host/apps modifications matter.
In any case, I am completely in the dark with regards to how serious
this problem is in real life. And frankly, this thread is not giving me
much new information in that regard. It could be that its not a problem
at all. Or it is. I would suggest that we take this issue seriously and
make sure there is someone making a measurement of what the actual
impacts would be. Data needed! We don't necessarily have to have a
complete new routing system to test or simulate this; it would be
sufficient to introduce the drop/delay-the-first-packet effect.
The other thing we should do is to go back to the tradeoff discussion
that David started; I found that useful. The questions that I have are:
1) Wouldn't a hybrid scheme be able to reduce the incidence of this
problem? 2) David's tradeoffs assumed that we operate either in push or
pull model. However, has it been established that we actually need a
separation architecture that needs these mappings? What about research
ideas such as compact routing?
to unsubscribe send a message to firstname.lastname@example.org with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg