Windy
The web crawler behind Semiote's search engine. If it has been to your site, this page is what the request pointed at.
- User agent
SemioteWindy/0.1 (+https://windy.semiote.org/crawler)- robots.txt token
SemioteWindy- Contact
- abuse@semiote.org
- Operated by
- Semiote
What it takes
Per site: robots.txt, the home page, and — only if the home page
looks like a site that has been maintained — pages linked from it or listed in
its sitemap, the stylesheets those pages reference, and one icon.
It never fetches JavaScript. Not a file, not a bundle, not a source map. There is no browser engine involved and nothing is executed. Script elements are removed from the markup before it is stored.
It also never requests images, video, fonts, archives or documents, never submits a form, never sends a cookie back, and never authenticates. If your site is not one it wants, it costs you two requests and does not come back for three months.
How to block it
Ordinary robots.txt, obeyed as
RFC 9309 describes it —
the longest matching rule wins and Allow breaks a tie.
User-agent: SemioteWindy
Disallow: /
Or part of it:
User-agent: SemioteWindy
Disallow: /private/
Allow: /private/press/
Naming Semiote instead blocks every crawler we will ever
run. A robots.txt that returns a server error is treated as a full
stop rather than as permission: if your server cannot state its rules, we leave
it alone until it can.
How to slow it down
User-agent: SemioteWindy
Crawl-delay: 10
Honoured. Without it we wait two seconds between requests and keep one
connection open at a time per host, however much work there is. A
429 or 503 stops us on the whole host, not just the one
address, and Retry-After is obeyed.
Who to write to
abuse@semiote.org, which is also sent as
the From header on every request, so it is in your access log beside
whatever prompted you to look. A person reads it.
How to check a request is really ours
Anyone can copy a user agent, and scrapers borrow reputable ones routinely. To confirm:
- Reverse-look up the address that connected. It resolves to a host under
crawl.semiote.org. - Forward-look up that host. It resolves back to the same address.
Both steps are needed; the second is what makes it unforgeable. Anything claiming to be Windy that fails this is not Windy, and we would like to know about it.