Caddy webserver goes south

teamacc

@caracal said:
Seems like they might have lost minio as a sponsor https://github.com/mholt/caddy/pull/1866

https://caddyserver.com/sponsor

Easy way to lose sponsors: "Hey sponsors, we kinda wanna do X, you ok with that?"

no response

"Ok, we're gonna do X"

WSS

My guess is that Minio checked out his Twitter, or may have read some of his other brilliance.

rm_

And that lost sponsorship was for $5000/year, worth of 50 "user licenses" for $100/year each -- and I bet he will not find 50 actual end-users to sign up and keep paying that -- ever -- for a webserver much slower than nginx, the only saving grace of which is "easy to set up for newbies" AND with the fully free fork (Wedge) widely known and available.

I.e. could have just stayed quiet and attract more corporate sponsors over time, but now the greed for more and now is causing a total failure.

WSS

Caddy will end up being a punchline to a nerd joke in XKCD. Countdown to how long began yesterday.

sarah

The change has been reverted now

lovespecialists

very good post i like it..

caracal

@sarah said:
The change has been reverted now

End of lowend-open source drama?

Rhys

@sarah said:
The change has been reverted now

The header was removed (because they lost a sponsor due to it, the irony), the restrictive EULA around the binaries still exists. Oh and matt is still a cock.

rincewind

Traefik looks much better than Caddy with some awesome features: LetsEncrypt, HTTP2, Hot reload, Docker/Kubernetes etc. They even got 1M in initial funding for opensource development. mholt just messed up with his funding strategy.

sarah

@Rhys said:

@sarah said:
The change has been reverted now

The header was removed (because they lost a sponsor due to it, the irony), the restrictive EULA around the binaries still exists. Oh and matt is still a cock.

Matt probably deserved losing a sponsor without communicating clearly to them but he hardly deserves to be name called/treated this way. He is just a grad student.

Rhys

@sarah said:

@Rhys said:

@sarah said:
The change has been reverted now

The header was removed (because they lost a sponsor due to it, the irony), the restrictive EULA around the binaries still exists. Oh and matt is still a cock.

Matt probably deserved losing a sponsor without communicating clearly to them but he hardly deserves to be name called/treated this way. He is just a grad student.

I have no sympathy for him, he is being a cock plain and simple.

[1]

https://twitter.com/mholt6/status/908544457980633089

and his snarky responses.

[2]

https://twitter.com/mholt6/status/908547005042982912

and what was posted before.

WebProject

typical example how to manage "open source" project (as I can see from a couple of projects now) :

1) create community of open source developers to create product "Caddy" web server

2) maintain and remove all the bugs + create various plugins

3) convert to commercial version when the products is perfect

4) restrict the usage and profit from it

Caddy
is
Open Source

haha, jokers

Aidan

Caddy appears to be rather shitty, I'm still trying to figure out a legitimate use for it - outside of being a special snowflake "unique".

Don't give me this 1-click-install BS either, there are many better performing options out there.

WSS

It's basically a solution for a problem that doesn't exist for anyone who is smart enough to not only code, but do basic system administration. Obviously Matt can't handle that.

MagicalTrain

@WSS said:
It's basically a solution for a problem that doesn't exist for anyone who is smart enough to not only code, but do basic system administration. Obviously Matt can't handle that.

Im lazy. I liked caddy. But what I liked about caddy the most was the configuration file. Made more sense to me than nginx.conf. Im switching back to apache2 as a backend and nginx/traeffic as a reverse proxy, though.

As all my services are basically for 20 people max, I never really cared about the performance.

WSS

@MagicalTrain said:
Im lazy. I liked caddy.

Well, there ya go.

@MagicalTrain said:

But what I liked about caddy the most was the configuration file.

As all my services are basically for 20 people max, I never really cared about the performance.

If anyone said they loved configuring Apache, they'd probably say the same about Sendmail.

But, ya know, after the initial pain, it isn't all that difficult.

I'm running Apache w/ PHP-FPM with mod_http2. For what I need, it's virtually instant, and the RAM use isn't bad.

There are plenty "Drop in and run" programs that did precisely what Caddy did, and had similarly-simple configuration. They tended to be Windows based with minimal CGI support, except for some goodies like fhttpd, which evidently got ported to Windows too.

MagicalTrain

@WSS said:
But, ya know, after the initial pain, it isn't all that difficult.

If it was actually all that bad, Id probably stay with caddy, even after this ordeal. But its mostly me being lazy. And with the drama the last two days, my laziness of not wanting to deal with the drama is more than the laziness of not wanting to configure apache/nginx.

WebProject

apache & nginx are better options anyway

eva2000

Caddy has reverted the HTTP change at least https://github.com/mholt/caddy/pull/1866. Though doesn't help with commercial licensing debate.

I've started building my own Caddy binaries from source now that the source builds have been fixed by Matt. It's part of Caddy integration into Centmin Mod I am testing privately right now. Plan to integrate a select few web servers including openlitespeed, litespeed and apache 2.4 and maybe h2o (not too fond of the syntax).

caddy -version
Caddy 0.10.9 (+545fa84 Fri Sep 15 14:36:05 UTC 2017)
1 file changed, 32 insertions(+)
caddy/caddymain/run.go

status

./caddy.sh status
caddy is running (13188)
-----------------------------------------------
Max processes     8192     8192     processes 
Max open files    262144   262144   files 
-----------------------------------------------

lol, I add a header within my Caddy integration installer for Centmin Mod but they're removal via Caddyfile vhost configs not forced

curl -I http://$(hostname):8888           
HTTP/1.1 200 OK
Accept-Ranges: bytes
Cache-Control: max-age=86400
Content-Length: 3801
Content-Type: text/html; charset=utf-8
Etag: "ow5yjt2xl"
Last-Modified: Tue, 12 Sep 2017 10:59:53 GMT
Server: Caddy
X-Powered-By: Caddy via CentminMod
Date: Fri, 15 Sep 2017 14:44:28 GMT

integrated plugins

caddy -plugins
Server types:
  http

Caddyfile loaders:
  short
  flag
  default

Other plugins:
  http.authz
  http.awses
  http.awslambda
  http.basicauth
  http.bind
  http.browse
  http.cache
  http.cgi
  http.cors
  http.datadog
  http.errors
  http.expires
  http.expvar
  http.ext
  http.fastcgi
  http.filter
  http.git
  http.gopkg
  http.grpc
  http.gzip
  http.header
  http.hugo
  http.index
  http.internal
  http.ipfilter
  http.jekyll
  http.jsonp
  http.jwt
  http.limits
  http.locale
  http.log
  http.login
  http.mailout
  http.markdown
  http.mime
  http.minify
  http.multipass
  http.nobots
  http.pprof
  http.prometheus
  http.proxy
  http.proxyprotocol
  http.push
  http.ratelimit
  http.realip
  http.reauth
  http.redir
  http.request_id
  http.restic
  http.rewrite
  http.root
  http.search
  http.status
  http.templates
  http.timeouts
  http.upload
  http.webdav
  http.websocket
  shutdown
  startup
  tls
  tls.storage.file

WSS

..does it still test at about 33% that of a stockish nginx?

eva2000

@WSS said:
..does it still test at about 33% that of a stockish nginx?

Haven't done proper testing with nghttp2's h2load HTTP/2 HTTPS load tester. But my statement is against my Centmin Mod built Nginx server, not stockish Nginx.

MikePT

Sounds good @eva2000, but remove your headers as well.

eva2000

MikePT said: Sounds good @eva2000, but remove your headers as well.

You mean for Nginx or Caddy ? Mine's just for Centmin Mod users and the header is user removable in both Centmin Mod Nginx and Centmin Mod built Caddy binary.

Did some quick benchmarks for siege 4.0.2 and wrk 4.0. Would have to move tests to non-openVZ server for nghttp2's h2load tests as my Docker image doesn't install on OpenVZ server which these tests are on https://community.centminmod.com/threads/caddy-http-2-server-benchmarks.5170/#post-54654

Hxxx

Thanks for the research / testing. I like these benchmarks.

@eva2000 said:

MikePT said: Sounds good @eva2000, but remove your headers as well.

You mean for Nginx or Caddy ? Mine's just for Centmin Mod users and the header is user removable in both Centmin Mod Nginx and Centmin Mod built Caddy binary.

Did some quick benchmarks for siege 4.0.2 and wrk 4.0. Would have to move tests to non-openVZ server for nghttp2's h2load tests as my Docker image doesn't install on OpenVZ server which these tests are on https://community.centminmod.com/threads/caddy-http-2-server-benchmarks.5170/#post-54654

MikePT

@eva2000 said:

MikePT said: Sounds good @eva2000, but remove your headers as well.

You mean for Nginx or Caddy ? Mine's just for Centmin Mod users and the header is user removable in both Centmin Mod Nginx and Centmin Mod built Caddy binary.

Did some quick benchmarks for siege 4.0.2 and wrk 4.0. Would have to move tests to non-openVZ server for nghttp2's h2load tests as my Docker image doesn't install on OpenVZ server which these tests are on https://community.centminmod.com/threads/caddy-http-2-server-benchmarks.5170/#post-54654

Both. I'd advise you not to take the same route, even if you allow it to be removed, I'd simply do not do it :-).

Other than that, your job has always been pretty awesome.

eva2000

MikePT said: Both. I'd advise you not to take the same route, even if you allow it to be removed, I'd simply do not do it :-).

Other than that, your job has always been pretty awesome.

Well my issue wasn't with the header being added, just the fact that for HTTP/2 based HTTPS performance, Caddy (and h2o) had a performance hit as you added more headers. If Caddy didn't have a performance hit, I wouldn't mind the added HTTP sponsor header. But this is first time using latest Caddy 0.10.9 with Go 1.9 so will have to recheck HTTP/2 HTTPS tests later.

a-super-random-user

Isn't Minio a free service of some sort?

MikePT

@sdglhm said:
Isn't Minio a free service of some sort?

It is, we're using it and it runs very well, their support @ github is really prompt as well.

@eva2000 said:

MikePT said: Both. I'd advise you not to take the same route, even if you allow it to be removed, I'd simply do not do it :-).

Other than that, your job has always been pretty awesome.

Well my issue wasn't with the header being added, just the fact that for HTTP/2 based HTTPS performance, Caddy (and h2o) had a performance hit as you added more headers. If Caddy didn't have a performance hit, I wouldn't mind the added HTTP sponsor header. But this is first time using latest Caddy 0.10.9 with Go 1.9 so will have to recheck HTTP/2 HTTPS tests later.

General issue here are the headers...

eva2000

MikePT said: General issue here are the headers...

Or that they're forced in Caddy official binaries ? cause every open source web server has an identifying server HTTP header as well - nginx, litespeed and apache. You can generally remove the identifying version number, but the web server itself is still shown in nginx and apache HTTP server header.

MikePT

@eva2000 said:

MikePT said: General issue here are the headers...

Or that they're forced in Caddy official binaries ? cause every open source web server has an identifying server HTTP header as well - nginx, litespeed and apache. You can generally remove the identifying version number, but the web server itself is still shown in nginx and apache HTTP server header.

They do, but still. Just an advice tho.