@inproceedings{walfish2004untangling,
   author =       "Michael Walfish and Hari Balakrishnan and Scott Shenker",
   title =        "{Untangling the Web from DNS}",
   booktitle =    {1st Symposium on Networked Systems Design and Implementation (NSDI)},
   year =         {2004},
   month =        {March},
   address =      {San Francisco, CA},
   abstract = {
The Web relies on the Domain Name System (DNS) to resolve the hostname portion of URLs into IP addresses. This marriage-of-convenience enabled the Web's meteoric rise, but the resulting entanglement is now hindering both infrastructuresnces. To this end, this paper describes the design and implementation of Semantic Free Referencing (SFR), a reference resolution infrastructure based on distributed hash tables (DHTs).
},
   notes = {
Suggests that, now DNS is commercial, "profit has replaced pragmatism as the dominant force shaping DNS ... Commercial pressures arising from its role in the Web have transformed DNS into a branding mechanism, a task for which it is ill-suited". The paper points out that URLs involving host names -- rather than services or objects -- makes some tasks unnecessarily difficult, like content replication. The paper suggests that web-references should be redesigned to be (i) persistent; and (ii) contention-free (have no relevance to trademark law for example). They suggest a system of opaque keys ("Semantic Free References") mapped by a DHT to concrete location records. In the proposed system, object keys must be found by search engines -- no more DNS name guessing -- and the implementation must handle network failures nicely (e.g. with DNS using the external network connection may still allow you to access internal names -- "fate-sharing")
<p>
Random notes (not intended to express any particular opinion):
<ul>
<li> I wonder how often people actually type in URLs manually... I usually find it easier to type keywords into google (&quot;I'm feeling lucky&quot;)... I don't really care what the URL bar says. When I want assurance I know to check for SSL anyway.
<li> Initiatives such as DOI are trying to create persistent object references for electronic publishing anyway
<li> The o-record (target of the keys in the DHT) look rather like mini HTML-redirection pages with caching hints and an onward pointer to either another reference or the final target
<li> This is the kind of thing someone like google could just do, particulary if they release their own browser
<li> They do remove the web URL misfeature where the host and path are exposed separately -- in this scheme, the full name is looked up in the name system <i>before</i> any HTTP request is generated. Good good.
</ul>.
Is this just another instance of &quot;every problem in Computer Science can be solved with another layer of indirection&quot;? Or just another combination of application and &quot;cool thing&quot; where the web is the application and DHTs are the &quot;cool things&quot; today?
},
   bibtexurl = {http://www.recoil.org/~djs/bibtex/untangling-web-from-dns.bib},
   pdf = {\url{http://http://nms.lcs.mit.edu/papers/sfr-nsdi04.pdf }},
}