W3C home > Mailing lists > Public > public-interledger@w3.org > November 2015

path discovery / routing: economic significance and brainstorming

From: Xavier Vas <xavier@tr80.com>
Date: Mon, 23 Nov 2015 13:31:35 +0000
To: public-interledger@w3.org
Message-ID: <56531523.7060808@tr80.com>

Anyone here agrees with me that path discovery, the thing that happens
before an ILP transaction as per white paper starts,
(a) will be what "defines" future inter-ledger payments in the sense
that its yet to be determined concepts and implementations will be what
makes or breaks the stated goals of inter-ledger payments (competitive
market with low fees, low friction user experience), much sooner than
concepts and implementations of the ILP itself, and
(b) is a hard problem.

As such it is surprisingly absent from both the white paper and the
"reference implementation", the latter which finds a path given a known
graph of who is connected to who, whereas in the real world finding
these connections is part of the problem.

If you agree join me for some path discovery brainstorming here on the
ML or, preferably, on IRC (irc://irc.w3.org:6667/#interledger) which is
sometimes a better forum for just tossing around half-baked ideas.

Just to throw some ideas out there:
- Someone suggested to look at inter-AS routing on the Internet for some
inspiration. In routing protocols like OSPF so-called "area border
routers" (ABRs) play a role similar to ILP connectors, connecting
subnets the way ILP connectors connect ledgers.
- Named data networking (NDN, http://named-data.net) has an as yet
mostly unsolved path discovery problem trying to retrieve data from a
network where it is not known which node(s) the data resides on. NDN's
"interest and response" scheme looks similar to the
preparation/execution cascade of ILP's universal mode. This could be
worth paying attention to.

Both ABRs and NDN scale by advertising "summaries" of nodes that can be
reached through them, in the ABR case through CIDR (common prefixes of
IP addresses) and in the NDN case through "path prefixes" which are
substrings of file system path-like URIs that they use to name content.
These "summaries" work because both IP addresses and file system paths
are hierarchical. How would a connector summarize ledgers/other
connectors that it connects to?

Food for thought.

Received on Monday, 23 November 2015 13:31:40 UTC

This archive was generated by hypermail 2.3.1 : Monday, 23 November 2015 13:31:41 UTC