Hacker News Reader

Comment by jimktrains2 | Hacker News Reader

This has nothing to do with gtld, but I think what bothers me most is that domains are in the wrong order. It should be com.ycombinator.news or com.google

View on HN

"I have to say that now I regret that the syntax is so clumsy. I would like http://www.example.com/foo/bar/baz to be just written http:com/example/foo/bar/baz where the client would figure out that www.example.com existed and was the server to contact. But it is too late now." Tim Berners Lee[1]

[1] http://www.w3.org/People/Berners-Lee/FAQ.html

Yes, now it is too late, since gTLDs have been rolled out in such a haphazard way. Right of first refusal on gTLDs should have been granted to the owner of the most highly trafficked domain (between .com, .org, .net, etc.), or worst case, been auctioned off between parties who already owned that domain on one of the TLDs. That would have meant the primary owner of a gTLD like 'google' would be google.com, instead of a random domain squatter.

Right of first refusal WAS given to trademark holders. .google is owned by Google, not by some random domain squatter (whatever gave you that idea?). Trademark vetting is part of the ICANN application process for a new gTLD. No one else other than Google would've been allowed to have the .google gTLD. There are plenty of brand name gTLDs that were never delegated simply because that company didn't want it, and no one else could have it.

Imagine how hard it would have been to untangle cross-site security if he had followed that scheme. As it is, we already have to have arbitrary restrictions on where in the domain hierarchy you can root things like cookies (e.g. if you're on www.example.com you can set cookies on example.com, but if you're on example.co.uk you can't set them on co.uk)

No worse than it is today, really. Just have a pubprefix instead of a pubsuffix list.

This is the single biggest UI mistake for security in browsers.

How do I explain to my grandmother when a banking site is safe? "Yeah, so, it has to start with 'https://', but only at the beginning, and then it has to END with "thebank.com", OR, before the third slash (so not the ones after https://). So find the third /: is there no third /? Then it must end with thebank.com. Is there one? Then before that it must have thebank.com. Yeah it's the part that's blacker! Exactly. It's not super clear but.... Ok forget it. Click the padlock and read the name of the company. And make sure it's from the same country, at least."

Tell her as I tell mine: The security behind the internet has at this point proven to be so fallible that it may as well not even be called security. The advice we I.T. folk have been preaching all these years [but quietly ignoring ourselves] is so much more profound than even we thought. You're safer to continue taking your bankbook to the same teller at the same bank you've been dealing with your whole life and continue to hope the bank doesn't get robbed while you're in there - because you're less likely to get killed in a bank robbery than you are crossing the road to get to the bank... both of which are infinitely less likely than your security getting compromised on the internet and your information being leaked to some hacker forum and ending up on the black market - regardless of the bank's laughable promise to you that your information is "safe and secure." It also has the additional upside of getting you out of the house, interacting with other people and keeping the human spirit alive instead of sitting cooped up inside all day.

In theory this is the problem that EV SSL Certificates solve.

[🔒 Bank Entity Inc.] https://yourbank.tld

That puts "com" first as you read it, despite the fact that it conveys no useful information.

Top-level domains have rarely been used as the ontological category they were intended to be. They are little more than flavor-text that is annoyingly required for uniqueness purposes. I think they rightfully belong in the position of least significance.

Sure, but "www" much less significant.

In an ideal world, 'www' would be superfluous. Your browser would know what server to contact for the WWW service by requesting SRV records for the bare domain.

> That puts "com" first as you read it, despite the fact that it conveys no useful information. They are little more than flavor-text that is annoyingly required for uniqueness purposes.

Except that's not true at all.

The reason they became "flavor-text" was because they appeared to be tacked on to the end for no reason other than uniqueness. Previously existing organizational schema worked for decades with proper categorization: Usenet is a wonderful example of just how powerful it is.

Had URLs been defined correctly, "com" would have immediately told the user "a commercial entity", "org" would have immediately meant "an organization", "net" probably wouldn't exist, and these newer TLDs like "google", "audio", "apps" would have made a hell of a lot more sense.

But nowadays TLD-s have little relation to the content category of the domain, look at the .com distortion for example. It's a valid claim that TLD-s are not really useful information. If you look at the current domain structure it's more like a file extension analogy.

The whole point is that this is only the case because of the TLD's position at the end of the root address, which has caused it to be perceived as a tacked-on additive for uniqueness' sake. If it had been placed at the beginning all along this likely would not be the case nowadays. TLDs very well could have been useful information; that's the point @awalton was making.

I agree to an extent, it would be interesting to see how the distribution of domains across TLD-s would look if it were reverse from the beginning, but I'm skeptical about a big shift. Country codes would still dominate local content and my feeling is .com / .net would be more balanced.

> Usenet is a wonderful example of just how powerful it is.

Usenet suffered from some a similar problem to domain names: everything started creeping into the "alt" top-level because it was the popular top-level free-for-all.

And the usefulness of the classification for humans was debatable since every topic could be found in multiple locations and some, like rec.arts.tv and alt.tv, rapidly ended up dwarfing entire top level categories like humanities.

Files, and the ordering of folder hierarchy from left to right, was developed by Unix in the 1970s.

Domain names (DNS) were developed in 1985 to locate things on the Internet, entirely separately, with the opposite hierarchical ordering.

The convention of locating files on a system identified by a domain name wasn't developed until 1992 and standardized in 1994 as a URL.

As a result, we have two hierarchical orderings for two separate things, because they were merged later on.

Note that postal addresses, at least in the United States, also order lines in descending specificity. I presume that long predates the Internet, and I wonder if it is in any way related.

What you describe is clearly familiar to any Java programmers out there, but it actually made me think of UUCP mail routing, which used to be a bit like that - the hosts would go from left to right, separated by bangs, and ending in the user. Some details here: http://www.tldp.org/LDP/nag/node192.html

What we have now is obviously much friendlier to the end user, given it has no reliance on remembering or understanding the network topology!

It's more like the old JANET domain names, that looked like UK.AC.MAN.CS.R5.

Ah yes, I remember those (dimly!)... though if you were sending mail to them it was still `user@` at the start. I don't remember when JANET reversed their ordering to match everyone else's - mid-nineties, maybe?

Wikipedia says it started hosting IP traffic in 1991, so that's probably when they started using DNS.

Maybe if you're a computer. But for us humans, google.com makes more sense.

> Maybe if you're a Java programmer.

Fixed that for you.

Why is it www.google.com and not google.com.www? And if www.google.com is okay, what's wrong with www.google/stuff. Left to right order.

Why is it www.google.com and not google.com.www

you know there's an actual reason for this?

com > google > www

"www" is a subdomain of "google" which is a subdomain of the top level domain "com"

historically that was a conventional way to separate services (ftp.google.com, www.google.com gopher.google.com)

Are you suggesting that the strategy you describe above wouldn't just as well if the names were reversed?

I think he is guessing that the grandparents thinks that www or 'world wide web' means 'the internet' and should be root.

But actually 'web' reveals to the graph of hyperlinks between http/html pages that let you navigate from page to page, server to server. www denotes the host is primarily a http server serving web pages.

You could argue it's superfluous these days - if you are typing it in an address bar, you hope it is a http server, and if you are typing it in a ssh prompt, you know it's your ssh server. People add it because it's a convention and some users might get confused if you don't... (it's probably a vhost anyway most of the time)

It does have one benefit. www.google.com has its IP served from google's dns servers, whereas google.com comes from com's. Google does lots of dns-based load balancing that com doesn't support.

That's not how it works.

com. only serves the nameserver (NS) addresses for example.com., and if they are within the same domain, their IP addresses. It does not serve the rest of the records for example.com.

You can see this easily with e.g.:

dig -t any google.com. @a.gtld-servers.net.

vs.

dig -t any google.com. @ns1.google.com.

There is no real advantage of having it be 'www' coming from DNS...

There is a somewhat more modern problem which is that having a domain be a CNAME (the DNS equivalent of a by-name redirect) means it can't also have MX records and receive mail; the CNAME also ends up "redirecting" everything, including MX lookups. So if your web host offers yourdomain.someawesomehost.com and you decide to point yourdomain.com to it via CNAME, having you@yourdomain.com doesn't work unless yourdomain.someawesomehost.com is configured to receive mail for yourdomain.com.

Reversed maintains the existing hierarchy but the suggestion was google.com.www which does not.

hostnames can have more than three components, e.g. "www.ox.ac.uk". So it's either biggest to smallest: "com.google.www" or smallest to biggest: "www.google.com".

google.com.www is in neither ascending or descending order.

www.google/stuff works now that we have generic TLDs and .google is a TLD, but before that it ignores the possibility of google.net and google.org. This may not be a problem for a unique trademark like google but not all domains are trademarks.

Really? com.ycombinator.news seems much more logical and intuitive to me, just like 2015-04-01.

Or how about com/ycombinator/news?

Because you're a computer programmer!

Human language is all about optimizing for context. Abbreviations, contractions, etc. For humans, most of the time the current year and/or "com" are the default context, so it makes sense to present the most significant differentiating information first. If you think about it, that's also why "www." got dropped.

When I make a phone call, I have a "default context" -- my area code -- but people still seem happy enough with phone numbers that go from most general to most specific.

I'm old enough to remember when people would drop area codes. Long distance was expensive so 95% of the numbers you'd call were local. That being the default context, people would drop the area code in most local situations. There's plenty of archival newspaper ads that demonstrate this.

With cell phones and unlimited nationwide calling plans, it's more common to find an assortment of area codes in contact lists these days. What you're describing is not actually a default context. It's just your personal area code. The person next to you might very well have a different one.

It's not my "personal area code", it's... well, the area's area code, and it's the default for people in the area. People and businesses locally routinely give their number without the area code, because it's implied and doesn't need to be dialed. Even my cellphone has the default context of the area in which the SIM card was bought.

So I'm not sure what "old enough to remember" means when it's still true today. 95% of the numbers I call now are local businesses, so I only dial seven digits.

I don't know, maybe you're from the future or something.

> it's the default for people in the area

It's only the default for new numbers. People carry their old numbers with them when they move, unlike the olden days. Bottom line, contact lists have a lot more mixed area codes now.

> 95% of the numbers I call now are local businesses

That might explain the difference of perspective here. Most people call other people on their phones, not just businesses.

At least according to the FCC, number portability was intended to allow moving between service providers. If you move across the country, there's no guarantee you'll be able to keep that number if you initiate service with another carrier. So in a sense, area codes are "supposed" to be meaningful (even if that's not so true today).

That's (English) linguistic bias. Not all human languages have the same conventions.

Only because you've gotten used to it. Right now you have to learn that [domain]/[filepath] is a thing.

If it was done in the same direction as folders, you wouldn't even have to know anything about domains.

Eh, that's debatable. Could be confirmation bias because we're just used to it being that way for the last x decades. Fairly sure that if I asked my computer illiterate grandma which made more sense, google.com or com.google, the answer would be "neither." Heck, my parents only add www or com simply "because". There's no rational reason from their perspective to do so other than that it may not work otherwise. If I had originally taught them com.google vs google.com, they'd be none the wiser (assuming either way actually worked).

I actually suspect if we started with a clean slate, it would make more sense to start with "more generic/less detail" to "less generic/more detail." Urls after the domain already do that (well, organized sites, do), e.g. www.site.com/cleaningsupplies/clorox

Then we could have (expanded for example): company.google organization.charitynavigator uk.company.google xxx.whoa (for those that remember the proposal for a xxx tld)

But there's probably a technical reason for it being the way it is that I'm not aware of or something.

Er... to me, if I really stop and think hard about it, none of these make sense. If I really wanted it to make sense, I'd just type "Google", "RBC", "BBC" - the brand name and it would figure out the rest for me. If the site was secure, my browser would figure that out and tell me it was secure. If the site wasn't secure, it would figure that out too and warn me as such.

Of course, I realize that this is a tall ask given that you have conflicting brands around the world. Perhaps there should be some linguistically obvious [to the average non-I.T. person] way to differentiate between conflicting brands...

I feel like the most obvious should be that we should never have to negotiate security by typing HTTPS... that should be the default state, falling back to HTTP only if HTTPS was unavailable and a warning should be given to the user. The average user [indeed any user] should never have to "look for HTTPS" or "look for the padlock". The whole browser UI should be more intuitive than that - the user should be able to look at the browser window and infer that communication with that site is not secure.

I'm not sure what the address structure should be... perhaps something like(?)

ISO_COUNTRY_CODE/REGISTERED_COMPANY_NAME/SITE_NAME/PATH/TO/RESOURCE

Examples:

- UK/COMPANY/ARGOS/ABOUT-US [Brand Information Page]

- UK/COMPANY/ARGOS/ONLINE-STORE/HOME/FURNITURE/BEDROOM/KIDS/BEDS/PIRATE-SHIP-BED [Catalog Item]

- UK/COMPANY/ARGOS/STORES/ACOCKS-GREEN [For local information]

- US/COMPANY/AMAZON [Brand Home Page]

- US/COMPANY/AMAZON/STORE/ELECTRONICS/DVD/LORD-OF-THE-RINGS [Catalog Item]

- US/CHARITY/LIVESTRONG [Brand Home Page]

- CA/CHURCH/UNITED/SASKATOON [Localized Home Page]

Perhaps a global brand such as Amazon or Facebook wouldn't even need a country code...or it could be figured out by an algorithm based on the user's context: FACEBOOK could map to UK/COMPANY/FACEBOOK for a person in the UK or US/COMPANY/FACEBOOK for a person in the U.S. The user could override the location by manually typing the address. If the user's IP address belonged to Amazon in the US, then they could type in just the site and resource address and the rest of the address could be construed from their context: Typing in STORE/ELECTRONICS/DVD/LORD-OF-THE-RINGS would automatically be understood as: US/COMPANY/AMAZON/STORE/ELECTRONICS/DVD/LORD-OF-THE-RINGS. For a residential user, their IP would be assigned to the block for their country, so any sites not explicitly specified would infer a site within their own country.

If the country code isn't known, then some algorithm could attempt to figure it out. If it was ambiguous, then the user could be presented with search results by a preconfigured [or their favorite] search engine.

Anyway - as Tim Berners Lee says... it's too late now. Until someone sets about re-imagining the entire infrastructure that is today's internet, we're stuck with what we've got.

https://en.wikipedia.org/wiki/X.400#Addressing

Everything old is new again.

Funny how history just keeps repeating itself ;) #IfItAintBroke

That's just because you are used to it. But <tld>.<domain> actually makes more sense for multiple reasons:

- That's how phone numbers work, the most significant part of the number is first (<country-code> <area-code> <local-number>) so the most significant part (TLD) should come first in domain names. Users like systems that look like other systems they are already used to.

- <tld>.<domain> would help thwart phishing attacks as most users read left-to-right, so they see right away when the domain is incorrect. Compare this to "google.com.phishing.com" which tricks users because they stop reading after they see "google.com..."

- When you read or hear "go to com.myawesomedomain" you know a domain name is being mentioned as soon as you hear "com.". It is not always obvious otherwise (you have to use specific language like "browse to").

How about US postal addresses?

Except for ZIP code, the address elements are ordered in descending specificity.

Addresses in Japan are written in exactly the opposite order, with the post code being written at the top of envelope followed by the prefecture, state, etc. Its simply what you are used to.

If it was com.something you have to type the com first. If "something" is first, you by the time you get to "som" it will probably have something.com in the dropdown list.

I would say its easiest if the most variable part is first, not only for autofill but for the way I think about it.

When you talk about "most significant" you mean the "largest" category. To me the actual most significant is the part that identifies the most unique thing you are typing in, and that tends to be the domain.

Modern web browsers will complete from any start position in an URL (and some also from the title of the document, I think). E.g. if I have visited the Wikipedia article on irish setters and I start typing irish, it will suggest that article. In the same fashion, if the URL was com.example and you typed exa, it should suggest com.example.

Saying the brand first definitely make more sense to mere mortals. Phone numbers not a good comparison because they are...numbers, not brand names.

Once again, you say that because you're used to it. When you go to a store, do you go to the item first, then figure out what section it's in then figure out what isle it's on?

No, but when you're looking for a particular website, you Goggle it, then check domain name is the one you were looking for, by having the name product/company name first, you don't need to scan the superfluous data.

That's true. I can definitely see your point. I think the long story, short here is to stick with what you're used to and by google having .google for domains leaves a bad taste in my mouth.

That's also how your filesystem works, and a little bit of similarity with something that is already well understood can only help.

arguably the .com is as superfluous as the now-omitted www.

It may be superfluous, but it has a dedicated button in my mobile phone, which helps me type less

why bother with .com at all - it is meant to be a global company but lots of us companies just use it for themselves (global or not) - just because the dns rules were made a long time ago doesn't mean we need to keep them today

Reply on news.ycombinator.com