the slash is a necessary evil. It delimits the steps in the instructions.

Also, most of the URL is where, things like HTTP, HTTPS, or FTP are the how. Its perfectly valid to put the "how" at either the very beginning or very end of a location set... but its long past time that web applications assume a reasonable default.

Wow, I think you just blew my mind a little bit. Where do I sign the petition?

Does that mean we can get started on a new comdot bubble?

that is possible... in the west, names are sometimes specific-to-general, like Product Marketing Manager... but other times its general-to-specific, like Vice President of Sales.

Likewise, addresses can go specific-to-general, as mentioned above. However, directions always go general-to-specific.

I think the problem may be that when DNS was first invented, it was for naming servers, not services. However, even as early as FTP, it became clear that the DNS name was just the first piece of a larger path...

With all their com.corp.utility.db.oracle class naming? I admit, the Java naming conventions I've seen (I'm not a developer, so I don't understand all the reasons) are pretty easy to understand and I suppose they organize fairly easily too.

P.S. Sorry for the late comment--just getting through my feeds after a hiatus.

The host-id (and optionally, the port) is a component of the URL, and may be a fully qualified domain name, a host name on the local network or an IP address. Figuring out what machine to forward the request to is the responsibility of the transport protocol (TCP) and the lower layers of the OSI network model (i.e. ethernet/WiFI/ARP and such). Network adapters, routers, switches and gateways do that stuff, and don't care about the application scheme (i.e. http or ftp) nor do they care about the url-path (i.e. /2008/06/dns-is-backwards). The naming convention for fully-qualified domain names was invented for the convenience of DNS long before URLs starting with "http" were relevant. Local network names are determined by the network OS (i.e. Novell or Active Directory, which both follow a specific-to-general format). Unlike FQ domain names and local network names, IP addresses follow a general-to-specific format as you parse the four octets left to right. Those things are like that for lots of reasons; making URLs roll off your tongue more naturally isn't one of them.
The url-path portion, on the other hand, is the purview of the application layers. They are interpreted by the web/ftp servers, browsers, ftp clients and such. The general-to-specific convention for url-path had different motivations behind it than the FQ domain name spec, and I suspect it's really a side effect of what's easier to parse in the programming languages that early implementations were written in.
Look at all the nonsense in the world:
Right now in my time zone it's 3:02 PM. Let's see, that's general (hour), specific (minutes), then more general (AM/PM means which half-day we're in). Military time would be better (15:02, with no ambiguity about half-day), but only soldiers, sailors and nerds use it, because apparently you need special training to "get it" viscerally.
What about dates? In the US, today is 07-17-2008, but in Europe it's 17.07.2008.
Don't get me started on the inability of Americans to deal with the metric system, or the fact that 7 days in a week makes no sense, or the bizarre collection of historical anecdotes that govern the number of days in a month.
My point is that the world is full of nonsensical crap like this, and URLs aren't the most egregious.