This post is completely trollish. The second comment in the thread explains the real meaning of the title extremely well. (Beeing really insightfull). The analogy for the Einsteins' title in the modern computer science will be f.e. "On the data distribution in the p2p networks", or "Stability of the Internet networks".
And these are the _real_ titles of the modern CS papers.
Check out my paper (to appear tomorrow at http://arxiv.org/list/cs/new - cs.NI/0205058).
From the abstract:
"We propose centralized algorithm of dat distribution in the unicast p2p network. Good example of such networks are meshes of WWW and FTP mirrors. Simulation of data propogation for different network topologies is performed and it is shown that proposed method performs up to 200% better then common apporaches".
The most funny thing about both of the technologies is that both of them are perfoming not as good as possible. For example you can have a look at http://www.terena.nl/conf/wcw/Proceedings/S4/S4-1. pdf - very nice paper that created a few hours long flame war between the authors and the Akamai people during the conference.
What is more interesting for the Open Source people is that Internet 2 team is working on Digital Storage Initiative (should be at http://dsi.internet2.edu/) that is creating the Content Delivery Network for the educational and research organizations. I'm sure that they have already that "secret sauce" so if you have any code (not words) you can try to convince them to work with theirs code.
The other funny thing about the Content Delivery Networks is that you can make them by yourself by integrating your own mirror in the Web Cache meshes (but this can be used mostly in Europe with a huge number of caches already installed).
The problem with AS records is that you actually can't create a good working mechanism only knowing the routing metric. You have to know something about the bandwidth awailable (that you can't), or use some indirect methods like measuring latency and throughoutput.
Akamai is inserting urls with some funny server names like http://a388.g.akamai.net/f/388/21/15m/www.cnn.com/ images/hub2000/ad.info.gif (for the cnn f.e.) that funny names are then looked up by your browser and the reply of the DNS server should point you to the list of the "nearest" Akamai servers.
Unfortunately, there is still no coherent architecture for mirroring
and for mirror sites to register their collections with the sites
which they mirror. In fact, we lack even a common (de facto) standard
for recording this replication information in a machine readable
for-mat. Some progress was made on this a few years ago by the
Internet Engineering Task Force s [1] working group on Internet
Anonymous FTP Archives, with the creation of the so-called IAFA
Templates [2]. These provided a simple machine readable format for
recording per-resource or collection metadata, which could easily be
created by hand or programatically. Although support for IAFA
templates was integrated into some software packages, e.g. the ALIWEB
search engine [3] and the ROADS resource discovery sys-tem [4] ,
this approach never became successful on a large scale. The World Wide
Web Consortium s Resource Description Format (RDF) [5] and the Dublin
Core metadata effort [6] may eventually provide a viable machine
readable interchange format.
Currently, the database underlying the freshmeat.net weblog [7] is
perhaps the closest thing we have to a genuine mirror registry -
though it focuses almost exclusively on soft-ware packages and
operating system distributions, and only offers limited mirror
informa-tion. RDF is also being used in this capacity as part of
rpmfind.net [8], although mirror information is very limited in this
case too. The Internet Engineering Task Force s Uni-form Resource
Names effort [9] is also relevant here, since it would be very useful
if there were persistent and location independent names for these
collections of replicated resources.
Another attempt to create a framework for such a metadata was an "Open-Software-Index" that Oliver Maruhn and myself tried to create almost 2 years ago. After this document some discussion had started (code name "Russian Freshmeat") that had shifted mostly to localisation of such a metadata. Unfortunately no working code was produced.
And at the end somewhat less relevant to the topic.
This kind of metadata should be extremely valuable for implementation of the URIs and particularly for the I2C(s) (URI tp URC). Quote from the RFC 2483:
"Uniform Resource Characteristics are descriptions of resources. This
request allows the client to obtain a description of the resource
identified by a URI, as opposed to the resource itself or simply the
resource's URLs. The description might be a bibliographic citation, a
digital signature, or a revision history. This memo does not specify
the content of any response to a URC request. That content is
expected to vary from one server to another."
Hopefully we already have mechanism for the I2L(s) (FTP Mirror Tracker).
Slashdot Mirror
This post is completely trollish. The second comment in the thread explains the real meaning of the title extremely well. (Beeing really insightfull). The analogy for the Einsteins' title in the modern computer science will be f.e.
"On the data distribution in the p2p networks", or
"Stability of the Internet networks".
And these are the _real_ titles of the modern CS papers.
Check out my paper (to appear tomorrow at http://arxiv.org/list/cs/new - cs.NI/0205058). From the abstract:
"We propose centralized algorithm of dat distribution in the unicast p2p network. Good example of such networks are meshes of WWW and FTP mirrors. Simulation of data propogation for different network topologies is performed and it is shown that proposed method performs up to 200% better then common apporaches".
The most funny thing about both of the technologies is that both of them are perfoming not as good as possible. For example you can have a look at http://www.terena.nl/conf/wcw/Proceedings/S4/S4-1. pdf - very nice paper that created a few hours long flame war between the authors and the Akamai people during the conference.
What is more interesting for the Open Source people is that Internet 2 team is working on Digital Storage Initiative (should be at http://dsi.internet2.edu/) that is creating the Content Delivery Network for the educational and research organizations. I'm sure that they have already that "secret sauce" so if you have any code (not words) you can try to convince them to work with theirs code.
The other funny thing about the Content Delivery Networks is that you can make them by yourself by integrating your own mirror in the Web Cache meshes (but this can be used mostly in Europe with a huge number of caches already installed).
The problem with AS records is that you actually can't create a good working mechanism only knowing the routing metric. You have to know something about the bandwidth awailable (that you can't), or use some indirect methods like measuring latency and throughoutput.
Akamai is inserting urls with some funny server names like http://a388.g.akamai.net/f/388/21/15m/www.cnn.com/ images/hub2000/ad.info.gif (for the cnn f.e.) that funny names are then looked up by your browser and the reply of the DNS server should point you to the list of the "nearest" Akamai servers.
And at the end somewhat less relevant to the topic.
This kind of metadata should be extremely valuable for implementation of the URIs and particularly for the I2C(s) (URI tp URC). Quote from the RFC 2483:
Hopefully we already have mechanism for the I2L(s) (FTP Mirror Tracker).Unfortunately the code is doing almost nothing and there is no progress on this for the last 2 years. So this not a solution.