I used to feel the same way. — Until the arrival of so many new TLDs.
Since then I always use www, because mentioning www.alice.band in a sentence is much more of a hint to a general audience as to what I’m referring to than just alice.band
I just redirect the other way round, so those ever rarer individuals typing in domains are also served fine on my websites. And also to automatically grab the https.
I just find it's ever slightly more "honest" to have the server name I mention, also be the one that's actually being served. -- And that's also because I'm quite annoyed at URL shorteners and all kinds of redirect trickery having being weaponized over the years.
So I optimize for honesty and facilitate convenience.
But this pretty subtle stuff and I'm not advocating anymore. -- I don't think it's that big of a deal either way and I'm just expressing my little personal vote and priorities on the big Internet. :-)
So my post wasn't intended to change your mind, but more as a bit of an alternative view and what made me get there.
Inbound email immediately broke. I was still very new, and didn’t want to prolong the downtime, so I reverted instead of troubleshooting.
A few months after I left, I sent an email to a former co-worker, my replacement, and got the same bounce message. I rang him up and verified that he had just set up the same firewall rule.
Been much too long to have any clue now what we did wrong.
From https://en.wikipedia.org/wiki/CNAME_record: "If a CNAME record is present at a node, no other data should be present; this ensures that the data for a canonical name and its aliases cannot be different."
So if you're looking up the MX record for domain, but happen to find a cname for domain to www.domain , it will follow that and won't find any MX records for www.domain.
The correct approach is to create a cname record from www.domain to domain, and have the A record (and MX and other records) on the apex.
Most DNS providers have a proprietary workaround to create dns-redirects on the apex (such as AWS Route53 Alias records) and serve them as A records, but those rarely play nice with external resources.
You can't do that, period.
A lot of "cloud" and other GUI interfaces deceive people into thinking it's possible, they just do A record fuckery behind the scenes (clever in it's own right but it causes misunderstanding).
When you have a www you ultimately have more flexibility. For example, you can point a CNAME at another CNAME. This answer on ServerFault mentions the additional options (and downsides of doing that): https://serverfault.com/a/223634 https://serverfault.com/questions/223560/www-a-record-vs-cna...
Heroku vaguely mentions the benefits under the "Limitations" section of this link: https://devcenter.heroku.com/articles/apex-domains
After a DDos attack, they were much more explicit in their recommendations: "We strongly recommend against using root domains. Use a subdomain that can be CNAME aliased to proxy.heroku.com, and avoid ever manually entering IPs into your DNS configuration." https://web.archive.org/web/20110609095616/https://status.he...
Here is an old post about someone who initially used an apex domain and then had issues (that they hacked around): https://web.archive.org/web/20110718170757/http://blog.y3xz....
I believe that some larger providers are providing some work arounds which makes it easier to hack around the issue these days, but I still firmly believe that if you set your site up using "www" (even if it is initially an A record - most of mine are A records right now), you will have more flexibility in the long run than if you set your site up on an apex domain.
I’m a no-www advocate. All my sites can be accessed from the Apex domain. But some people for whatever reason like to prepend www to my domains, so I wrote a rule in Apache’s .HTACCESS to rewrite the www to the Apex.
Here’s a tutorial for doing that: https://techstream.org/Web-Development/HTACCESS/WWW-to-Non-W...