Is there any performance difference between workers and snippets?

W

Walshy•6/22/23, 5:47 PM

I'd have to test it to be 100% sure but it should be fine

PPing for toast Is there any performance difference between workers and snippets?

W

Walshy•6/22/23, 5:47 PM

Snippets are just Workers so no

WWalshy Cache doesn't count towards a subrequest afaik

K

kian•6/22/23, 5:49 PM

It counts towards the 50 on a normal Worker meowthinking

K

kian•6/22/23, 5:49 PM

Is snippets special?

P

Ping for toastOP•6/22/23, 5:49 PM

owch. let's see if we can trim this down

P

Ping for toastOP•6/22/23, 5:50 PM

wow pages with full SSR are only a couple ms slower than this simple worker... I have some optimization to do!

Kkian It counts towards the 50 on a normal Worker <:meowthinking:350680941804388352>

W

Walshy•6/22/23, 5:51 PM

oh it does? damn

W

Walshy•6/22/23, 5:51 PM

nevermind then

K

kian•6/22/23, 5:51 PM

it’s not supposed to but they also didnt plan to fix it since “50 is enough”

K

kian•6/22/23, 5:51 PM

W

Walshy•6/22/23, 5:51 PM

the 1 subrequest limit is a little awkward ablobsweats

Kkian it’s not supposed to but they also didnt plan to fix it since “50 is enough”

W

Walshy•6/22/23, 5:52 PM

ah...

W

Walshy•6/22/23, 5:52 PM

that makes sense

W

Walshy•6/22/23, 5:52 PM

yeah we should probably fix that now for Snippets

K

kian•6/22/23, 5:52 PM

well, fix as in move it to be inhouse subreq like KV/etc

P

Ping for toastOP•6/22/23, 5:53 PM

is there any detailed docs about the expensive parts of HTMLRewriter? It would be interesting to know what's consuming the largest amount of CPU time

K

kian•6/22/23, 5:53 PM

kj::Own<CacheClient> IoContext::getCacheClient() {
  // TODO(someday): Should Cache API requests be considered in-house? They are already not counted
  //   as subrequests in metrics and logs (like in-house requests aren't), but historically the
  //   subrequest limit still applied. Since I can't currently think of a use case for more than 50
  //   cache API requests per request, I'm leaving it as-is for now.
  limitEnforcer->newSubrequest(false);
  return getIoChannelFactory().getCache();
}

kj::Own<CacheClient> IoContext::getCacheClient() {
  // TODO(someday): Should Cache API requests be considered in-house? They are already not counted
  //   as subrequests in metrics and logs (like in-house requests aren't), but historically the
  //   subrequest limit still applied. Since I can't currently think of a use case for more than 50
  //   cache API requests per request, I'm leaving it as-is for now.
  limitEnforcer->newSubrequest(false);
  return getIoChannelFactory().getCache();
}

kj::Own<CacheClient> IoContext::getCacheClient() {
  // TODO(someday): Should Cache API requests be considered in-house? They are already not counted
  //   as subrequests in metrics and logs (like in-house requests aren't), but historically the
  //   subrequest limit still applied. Since I can't currently think of a use case for more than 50
  //   cache API requests per request, I'm leaving it as-is for now.
  limitEnforcer->newSubrequest(false);
  return getIoChannelFactory().getCache();
}

kj::Own<CacheClient> IoContext::getCacheClient() {
  // TODO(someday): Should Cache API requests be considered in-house? They are already not counted
  //   as subrequests in metrics and logs (like in-house requests aren't), but historically the
  //   subrequest limit still applied. Since I can't currently think of a use case for more than 50
  //   cache API requests per request, I'm leaving it as-is for now.
  limitEnforcer->newSubrequest(false);
  return getIoChannelFactory().getCache();
}

W

Walshy•6/22/23, 5:53 PM

that says

false

false

false

false though

K

kian•6/22/23, 5:54 PM

because that param is

isInHouse

isInHouse

isInHouse

isInHouse

W

Walshy•6/22/23, 5:54 PM

I guess the newSubrequest is what actually does it

K

kian•6/22/23, 5:54 PM

W

Walshy•6/22/23, 5:54 PM

ah

K

kian•6/22/23, 5:54 PM

virtual void newSubrequest(bool isInHouse) = 0;

virtual void newSubrequest(bool isInHouse) = 0;

virtual void newSubrequest(bool isInHouse) = 0;

virtual void newSubrequest(bool isInHouse) = 0;

W

Walshy•6/22/23, 5:54 PM

want to PR Kian? LuL

C

Chaika•6/22/23, 5:54 PM

You can still use cache API with Snippets, just check cache, if it's not there, redirect with a query parameter to not check it /s

W

Walshy•6/22/23, 5:54 PM

or i guess make an issue/discussion first and see what Kenton says

CChaika You can still use cache API with Snippets, just check cache, if it's not there, ...

W

Walshy•6/22/23, 5:55 PM

pls no

K

kian•6/22/23, 5:55 PM

sgtm

K

kian•6/22/23, 5:55 PM

ship it

Kkian ```cpp kj::Own<CacheClient> IoContext::getCacheClient() { // TODO(someday): Sh...

P

Ping for toastOP•6/22/23, 5:55 PM

https://github.com/cloudflare/workerd/blob/e1799c742439602804b9a38cc1c01a9bff796bb7/src/workerd/io/io-context.c%2B%2B#L910-L917

B

bkyerv•6/22/23, 6:02 PM

is cloudflare going to launch user management service like clerk dev or supabase auth?

W

Walshy•6/22/23, 6:09 PM

Yeah it'd be great to have AE for sure

S

sakuta•6/22/23, 6:09 PM

how effective is 1.1.1.1's malware blocking? is the list updated daily, hourly? what sources does it take them from?

MM 1.1.1.1 doesn't block malware, you want 1.1.1.2

D

dave•6/22/23, 6:12 PM

huh neat, didn't know that was a thing

MM 1.1.1.1 doesn't block malware, you want 1.1.1.2

S

sakuta•6/22/23, 6:16 PM

yeah sure, but it's a subsidiary of 1.1.1.1

D

dave•6/22/23, 6:20 PM

nice, works in Chrome too!

MM 1.1.1.1 doesn't block malware, you want 1.1.1.2

C

Chaika•6/22/23, 6:21 PM

lol automod does not like the word you got timed out for an hour for, .3 is for nsfw blocking

S

sakuta•6/22/23, 6:27 PM

how accurate is it tho?

Ssakuta how accurate is it tho?

D

dave•6/22/23, 6:30 PM

I'll let ya know next time North Korea targets me.

Ssakuta how effective is 1.1.1.1's malware blocking? is the list updated daily, hourly? ...

C

Chaika•6/22/23, 6:30 PM

To answer your question, all Radar FAQ says is:

"Cloudflare uses a variety of data sources to categorize domains."

You can use the domain feedback form to suggest something be recategorized: https://radar.cloudflare.com/domains/feedback
You can also look up domains there, in my limited experience it's fairly accurate. If you use full Zero Trust, you get a custom DoH/DoT address you can use, and then you can create a policy to block whatever categories, and then make any exclusions as necessary for false postives

S

sakuta•6/22/23, 6:33 PM

So it's probably better than the malware blocking of some of the VPNs out there

S

sakuta•6/22/23, 6:33 PM

Good to know!

Ssakuta So it's probably better than the malware blocking of some of the VPNs out there

C

Chaika•6/22/23, 6:37 PM

ahh you asked specifically about Malware blocking, got distracted by Synthetic getting automod timed out for an hour when he mentioned .3 blocking nsfw and thought you were asking about that lol.
I don't know how correct it is about malware. Overall, with nsfw/etc sites, the categories seem right most of the time. I've only dealt with it false blocking things, it does seem fairly fast about picking data up from those malware reporting websites, not sure which one it uses as sourrces. One guy got his site reported on one of those, and he was on Radar like ~2 days later complaining about it blocking his site
With ZT there is a number of other options you can pick as well, like blocking brand new sites, newly seen domains, etc.

C

Chaika•6/22/23, 6:45 PM

lol that sounds like a them issue to be fair, depending on how dodgy the sites are

C

Chaika•6/22/23, 7:16 PM

Interactive Challenge?

Is there any performance difference between workers and snippets?

Similar Threads

Similar Threads

Similar Threads