256

Other than for historical reasons, is there is reason to have “www” in a URL?

Should I create a permanent redirect from www.xyz.com to xyz.com, or from xyz.com to www.xyz.com? Which one would you suggest and why?

Chad
  • 113
  • 4
Quintin Par
  • 4,293
  • 10
  • 46
  • 72

7 Answers7

213

One of the reasons why you need www or some other subdomain has to do with a quirk of DNS and the CNAME record.

Suppose for the purposes of this example that you are running a big site and contract out hosting to a CDN (Content Distribution Network) such as Akamai. What you typically do is set up the DNS record for your site as a CNAME to some akamai.com address. This gives the CDN the opportunity to supply an IP address that is close to the browser (in geographic or network terms). If you used an A record on your site, then you would not be able to offer this flexibility.

The quirk of the DNS is that if you have a CNAME record for a host name, you cannot have any other records for that same host. However, your top level domain example.com usually must have an NS and SOA record. Therefore, you cannot also add a CNAME record for example.com.

The use of www.example.com gives you the opportunity to use a CNAME for www that points to your CDN, while leaving the required NS and SOA records on example.com. The example.com record will usually also have an A record to point to a host that will redirect to www.example.com using an HTTP redirect.

Greg Hewgill
  • 6,749
  • 3
  • 29
  • 26
  • 5
    You can provide a "default" CNAME record pointing to the CDN, you don't have to use "www". This allows your DNS server to have SOA, NS, CNAME, etc RRs all for the same domain name. – Chris S Oct 28 '10 at 01:19
  • 3
    How come nobody mentions the `ALIAS` (or `ANAME` records) in this kind of topic? Doesn't it achieve the same results as the CNAME on nakeddomain (except from the cookie issue...)? – Augustin Riedinger Feb 27 '14 at 10:44
  • 3
    @AugustinRiedinger: ANAME records are not a standard DNS RR type. They are proprietary to specific service providers. – Greg Hewgill Feb 27 '14 at 16:57
  • But does this generates any compatibility issue or something? Is there any reason we should not use them (besides not being standard but proprietary)? – Augustin Riedinger Feb 27 '14 at 20:21
  • Not ONLY a quirk - it also limits cookies to not be send to every machine but only those in the www domain. – TomTom Mar 17 '14 at 13:50
  • 2
    @AugustinRiedinger it is not supported on most DNS **servers**. But if your **provider** has a DNS server which supports these features then that shouldn't give any problems with clients. – Koen. Jul 18 '14 at 16:02
107

Note: As of the ratification and implementation (by all current browsers, except possibly MSIE 11, see comments) of RFC 6265 in 2011 the following is no longer accurate, since cookies are by default never set across subdomains.

Historically, one good technical reason to make www.example.com canonical was that cookies of a main domain (i.e. example.com) were sent to all subdomains.

So if your site used cookies, they would be sent to all its subdomains.

Now, this often makes sense but it’s positively harmful if you only want to download static resources since it just wastes bandwidth. Consider all the style sheets and images on your website: usually, there’s no reason to send cookies to the server when requesting an image resource.

A good solution is therefore to use a subdomain for static resources, such as static.example.com, to save bandwidth by not sending cookies. All images and other static downloads can be downloaded from there. If you now use www.example.com for the dynamic content, this means that cookies only have to be sent to www.example.com, not to static.example.com.

If, however, example.com is your main site, then cookies will be sent to all subdomains, including static.example.com.

Now this isn’t relevant for most sites but changing your canonical URL later on isn’t a good idea so once you settled for example.com instead of www.*, you’re basically stuck with it.

An alternative is to use a whole different URL for static resources. Stack Overflow for example uses sstatic.net, YouTube uses ytimg.com etc. …

Konrad Rudolph
  • 1,286
  • 1
  • 9
  • 11
  • 11
    By the way, I really don’t like `www.x` as the canonical URL so personally I would probably use a different URL for static resources if I were to design a big site. – Konrad Rudolph May 27 '10 at 08:19
  • Helps with CDN and setting up media.example.com, et al. to use proper subdomains. The cookie thing causes extreme green around the gills and bloating when it rears its ugly head. – Fiasco Labs Sep 15 '13 at 20:36
  • This answer is entirely wrong in every important way. *Neither* cookies created on the server or with JavaScript on the apex domain (example.com) are shared with subdomains (www.example.com) by default, and vice-versa. However, both the main domain and subdomains *can choose* to create cookies with `domain=.example.com`, in which case they will be shared with the apex domain and subdomains. Therefore, cookies make *absolutely no difference* to the question of whether to use www or not. – Robin Winslow Jul 11 '18 at 13:56
  • 1
    @RobinWinslow But setting cookes on `domain=example.com` will set the cookie on the apex domain *and on subdomains*, and a way to avoid this is to not use the apex domain for HTTP. Though, agreed, another way would be to simply not specify `domain` when setting the cookie. I wonder if this has changed since I wrote my answer (which predates the relevant [RFC 6265](https://tools.ietf.org/html/rfc6265)!) but I can’t be bothered to look it up now. – Konrad Rudolph Jul 11 '18 at 14:15
  • @KonradRudolph yes I think what's happened here is a change in how modern browsers handle cookies. I'm trying to track it down. But the important point is that the *default* if you simply set cookies on the apex domain, in latest Firefox and Chromium, is that they won't be shared with subdomains. Maybe in older browsers they still will be? I'd like to find a definitive source. – Robin Winslow Jul 11 '18 at 14:44
  • Assuming it is a change in standard practice then my comment was unfairly critical, I apologise. But it appears that cookies are *no longer* a reason to choose www or not, in the vast majority of browsers. Although it would be nice to be able to put exact percentages to that statement. – Robin Winslow Jul 11 '18 at 14:46
  • 3
    It looks like the behaviour I describe has been the case since at least 2011 when RFC 6265 was written (more as a summary of current browser behaviour than a statement of how they should work). By now, we can assume that all browsers will be following it. See https://stackoverflow.com/questions/1062963/how-do-browser-cookie-domains-work and http://bayou.io/draft/cookie.domain.html. Given this, I think your answer has been misleading for at least 7 years, although it *could have* been accurate in some cases at the time of writing. Could you please update it to clarify this fact? – Robin Winslow Jul 11 '18 at 14:58
  • 3
    @RobinWinslow Yes, will do. – Konrad Rudolph Jul 11 '18 at 15:36
  • 1
    @KonradRudolph actually, according to https://www.mxsasha.eu/blog/2014/03/04/definitive-guide-to-cookie-domains/, the undesirable behaviour you described may have existed in IE11 at the time that post was written, and may still - which remains the most recent version of Internet Explorer. That would be pretty significant and worth mentioning, but it would be good to check it first. I can't check easily, as I'm on Ubuntu, but if you could that would be wonderful. – Robin Winslow Jul 11 '18 at 15:46
  • 1
    @RobinWinslow Ah, unfortunately I can’t either. :/ I will just mention the caveat. – Konrad Rudolph Jul 11 '18 at 15:48
  • 2
    More evidence that IE11 does the bad thing: https://developer.microsoft.com/en-us/microsoft-edge/platform/issues/8183708/#comment-5 – Robin Winslow Jul 11 '18 at 15:50
14

www is a subdomain usually used for the web server on a domain along with others for other purposes such as mail etc. Nowadays, the subdomain paradigm is unnecessary; if you connect to a website in a browser, you'll get the website, or sending mail to the server will use its mail service.

Using www or not is a matter of personal preference. Opposing points of view can be found at http://no-www.org/ and http://www.yes-www.org/ - however, I believe that www is unnecessary and just adds more cruft to the URI.

Most servers send the same site either way, but don't redirect. For SEO purposes, choose one, then get the other to redirect to it. For example, some PHP code to do this:

if (preg_match('/www/', $_SERVER['SERVER_NAME'])) {
  header("Location: http://azabani.com{$_SERVER['REQUEST_URI']}");
  exit;
}

However, some reasons promoting the use of a www subdomain made by other answerers are great too, such as not sending cookies to static servers (credit Konrad Rudolph).

Delan Azabani
  • 528
  • 1
  • 4
  • 10
  • 3
    Looks like http://no-www.org/ has reverted to a for-sale parked page where http://www.yes-www.org/ is still going strong. I guess that settles it. Everybody use "www" from now on. – nunya Jan 05 '18 at 01:07
8

If you are going to have subdomains for other purposes (blog for instance), you may want to differentiate the sites and have a www prefix for the regular site. Other then that, the only important thing is to pick one of the two and stick to it (for SEO reasons).

Oded
  • 267
  • 2
  • 16
  • 1
    I can't find reference at the moment, but it could also have effects on the same origin policy. – Kobi May 27 '10 at 08:09
  • Yes it will, unfortunately. You can't AJAX to `www.example.com` from `example.com` or vice versa without something like JSONP. – Delan Azabani May 27 '10 at 08:12
8

It's pretty historical. Once upon a time we used to have www.example.com, ftp.example.com, images.example.com, uk.example.com etc which seemed like a logical thing to do and provided a simple method to spread the load between servers.

These days I would just go for example.com for the main site and redirect the www version to that.

The Google Webmaster tools allow you to specify your preferred domain, so make sure you use those too.

See also:
https://stackoverflow.com/questions/1109356/www-or-not-www-what-to-choose-as-primary-site-name
https://stackoverflow.com/questions/1884157/to-www-or-not-to-www

7

I'd do the first. The www convention comes from the early days of HTTP where www.cmu.edu and cmu.edu were very likely different machines.

msw
  • 399
  • 1
  • 5
  • 10
    in the 'early days' you'd rarely see an A record for a domain -- maybe it'd have an MX record, but you rarely had a host there. – Joe H. May 27 '10 at 12:27
3

Here is another minor perspective.

By not having www, there is a minor downside when it comes to text-based media, whether printed or online, and that is getting it recognized as a web address. In print, it’s usually pretty obvious that example.com is a web address, and you can add styling touches to highlight that. But plain text online? Not so easy. The chances are that if you send a plain text message – whether email, tweet, Facebook post, SMS or whatever – it will recognize a URL beginning with http:// or www. but won’t recognize one without either of those. So in order to make the URL into a clickable link, you have to either put www. or http:// on the front, and of the two, www. is shorter, less clunky to look at and easier to read.

Ray
  • 159
  • 4
  • 4
    `http://example.com/` is fully qualified whereas `www.example.com` is not. I prefer the fully qualified approach because it is _always_ recognisable as a URL regardless of whether it's `https://example.uk/` or `https://blog.example.eu/` or whatever. It is consistent with specifying the protocol of a secure site as HTTPS; `www.example.com` is just a domain and says nothing about which protocol one should use to access it. – James Haigh Mar 17 '15 at 01:11