Rate limiting is very basic about meg

[tomnomnom]

@hrbrmstr I've added some basic rate limiting in fffe252 - and a warning about reducing it in the readme.

'basic' might be a bit of an understatement, but it's an improvement. I'll give some more thought to it over the weekend.

I'm going to leave this issue open until I'm comfortable about the solution - maybe even until it fetches, parses and respects the robots.txt.

Thanks again!

[tomnomnom]

Thanks for the heads up and super detailed reply!

I'm working on rate limiting at the moment. The tool has been for my own use up until now and I've been putting it off - largely because I've been trying to come up with some kind of requeuing mechanism to keep the pipeline of requests as full as possible and avoid stalling goroutines. Given your input I think I'll implement something simpler but less efficient as an interim measure.

Do you think default 5s delay with an override would be OK for now? I think I'm a bit away from fetching and parsing 🤖 yet.

WRT the user agent: yeah; that's a leftover from the only-for-my-use days. I've dropped that in 4418a31. I'll add a 'meg' mozilla-like user agent soon and give people the option of overriding if they want to.

WRT HAR format: I'll look in to that. The original idea for this tool was to make it easy for me to use standard tools like grep to filter and find things in the output. Is HAR fairly amenable to that?

[hrbrmstr]

w/r/t HAR: aye. It's lovely JSON, so you can even use jq to do cmd-line filtering, querying and selecting. Go has some nice HAR tools as does R and Python, too. Def not a "must have" and WARC is far more "grep" friendly.

[hrbrmstr]

ref: https://twitter.com/hrbrmstr/status/947064265621549058

This is a 👍 issue to bring up.

Given that you're directing this at bug bounty hunters, the per-org bug bounty rules of engagement need to state that it's OK to ignore Disallow rules in robots.txt (using 🤖 from now on for that file ref). Why? LinkedIn — but many others in 2017 — have successfully punished (civil court) ppl and orgs for ignoring it and there are a cpl suits on the bench that could become case law starting in 2018 which formalizes penalties. Verbiage when running it or inherent checks & force of a -- switch to override some default check & halting behaviour would help keep the community and you as a tool author from being at-risk.

🤖 has optional Crawl-Delay settings in it. If not present the common standard is 5s and the kind standard is 10s between requests to same origin domain resources. The same came suits that are likely becoming case law in 2018 also used ignoring this as a reason for civil damages recompense. Unless the bug bounty program says it's OK to ignore this, a similar default behavior with forced manual override should be considered.

Slightly on-topic is

meg/http.go

Line 37 in 29116f8

// It feels super nasty doing this, but some sites act differently

. I scan the internet as part of $DAYJOB and totally understand where you're coming from. However you've essentially built a bot and default behavior (this is all to avoid potential penalties) is to have your own UA. Said UA can look like a "Mozilla"-ish string but it should have your bot into in it. Enabling another -- switch to override this and use one provided by the user OR to short-name select from an internal select list of canned ones wld be safer for users and you. https://github.com/hrbrmstr/splashr/blob/a7c5406264b91918e60e5abf692b51baf5ab2fb7/R/dsl.r#L422-L468 has a good set of ones to use for that purpose AND the added benefit is that folks can switch OS and from desktop to mobile sine many sites change behaviour with different UAs present.

Completely off-topic: you're capturing tons of juicy data. Consider using HAR format for the output (https://github.com/CyrusBiotechnology/go-har may help with that). It's slightly better than WARC (IMO) for storing some technical info along with the headers and payload but WARC support might also be nice (https://github.com/slyrz/warc may help) since there are at-scale tools to enable working with that.

[hrbrmstr]

If I can carve out some cycles in wk2 of Jan (I'm on a pretty big project until then) I'll gladly lend a hand. This is a super nice and focused alternative to generating URL strings for wget (which can do much of this but not with the 'cyber' eye and purpose you have).

[hrbrmstr]

(replying as I read, apologies :-)

def 5s is — IMO — 100% 👍 https://rud.is/b/2017/07/28/analyzing-wait-delay-settings-in-common-crawl-robots-txt-data-with-r/

[tomnomnom]

@hrbrmstr superb; thank you so much :)

I'll definitely look more into HAR in that case (although I may use gron for making it easier to grep - if you'll forgive the shameless self promotion :-P).