lawa
lawa is my web crawler. It crawls the web and writes down the HTTP headers that come back from each site. Everything below is regenerated every few minutes from what it has seen so far.
Status
| lawa is | crawling |
|---|---|
| pace | 1,284 requests in the last minute |
| servers visited | 7,148,514 finished, 750 in progress |
| servers waiting in line | 9,163,816 |
| responses studied below | 83,249,542 from 6,202,513 servers |
| numbers as of | 2026-11-05 10:20 UTC |
The shares below are measured by hostname. Additionally, lawa takes at most 25 pages from any one server, and each server is counted once, when lawa has finished with it.
Web servers seen
| Server header | servers | share |
|---|---|---|
| cloudflare | 1,621,159 | 26.1% |
| apache | 1,126,651 | 18.2% |
| nginx | 1,073,234 | 17.3% |
| (none) | 515,081 | 8.3% |
| gse | 232,397 | 3.7% |
| openresty | 221,139 | 3.6% |
| squarespace | 220,280 | 3.6% |
| litespeed | 159,452 | 2.6% |
| github.com | 116,715 | 1.9% |
| vercel | 97,656 | 1.6% |
13.7% of the servers that name themselves also announce their version number.
| X-Powered-By | servers | share of those sending it |
|---|---|---|
| php 8 | 259,712 | 36.8% |
| php 7 | 88,306 | 12.5% |
| asp.net | 87,598 | 12.4% |
| next.js | 59,200 | 8.4% |
| php 5 | 30,233 | 4.3% |
| plesklin | 29,332 | 4.2% |
| express | 26,971 | 3.8% |
| wp engine | 22,715 | 3.2% |
Bad time
A Date only has one‐second resolution and the network adds delay, so I decided that “on time” means “within 2 seconds, from my perspective”.
| server clock | servers | share |
|---|---|---|
| on time | 5,451,628 | 98.6% |
| up to 10 seconds off | 21,946 | 0.4% |
| 10 seconds to a minute | 18,300 | 0.3% |
| a minute to an hour | 35,254 | 0.6% |
| an hour to a day | 3,308 | 0.1% |
| more than a day | 166 | 0.0% |
Of the 78,974 wrong clocks, 53,827 run slow and 25,147 run fast. 1,368 are wrong by a whole number of hours, which is likely timezone problems rather than drift. 48,535 servers sent no Date at all, and 884 sent malformed Dates.
Responses served from a cache (indicated by an Age header) are left out. 620,689 servers only ever answered that way.
Security headers
| header | servers | share |
|---|---|---|
| HSTS (share of HTTPS servers) | 1,943,451 | 38.6% |
| X-Content-Type-Options | 2,713,098 | 43.7% |
| X-Frame-Options | 1,731,365 | 27.9% |
| Content-Security-Policy | 1,088,583 | 17.6% |
| Referrer-Policy | 1,037,592 | 16.7% |
| Permissions-Policy | 496,146 | 8.0% |
| X-XSS-Protection (deprecated) | 1,072,421 | 17.3% |
TLS
81.2% of servers answered over HTTPS at least once. lawa connects even when the certificate is bad, so here’s what was wrong. 2.1% of HTTPS servers would have failed in a browser.
| certificate problem | servers | share of HTTPS |
|---|---|---|
| wrong hostname | 59,952 | 1.2% |
| unknown or incomplete issuer | 25,664 | 0.5% |
| expired | 25,498 | 0.5% |
| self‐signed | 11,633 | 0.2% |
| protocol | servers | share of HTTPS |
|---|---|---|
| TLSv1.3 | 4,697,497 | 93.3% |
| TLSv1.2 | 336,700 | 6.7% |
| certificate authority | servers | share of HTTPS |
|---|---|---|
| Let's Encrypt | 2,864,495 | 56.9% |
| Google Trust Services | 1,238,038 | 24.6% |
| Sectigo Limited | 208,760 | 4.1% |
| Amazon | 172,242 | 3.4% |
| DigiCert Inc | 164,039 | 3.3% |
| GoDaddy.com | 78,517 | 1.6% |
| GlobalSign nv-sa | 61,982 | 1.2% |
| ZeroSSL GmbH | 22,475 | 0.4% |
Other quirks
lawa speaks my handrolled HTTP/1.1 implementation and logs each header block as the raw bytes, so we can see some stuff that HTTP libraries usually clean up.
| quirk | servers | share |
|---|---|---|
| same header sent more than once | 1,284,994 | 20.7% |
| every header name in lowercase | 334,088 | 5.4% |
| answered HTTP/1.1 with HTTP/1.0 | 37,723 | 0.6% |
| bare LF line endings (no CR) | 714 | 0.0% |
| folded (multi‐line) header values | 403 | 0.0% |
The average response carried 15.2 header lines in 818 bytes. In total, lawa has read 68.09 GB of headers.
Status codes
| status | responses | share |
|---|---|---|
| 200 | 63,552,249 | 84.8% |
| 301 | 6,130,199 | 8.2% |
| 404 | 1,766,884 | 2.4% |
| 302 | 1,746,484 | 2.3% |
| 403 | 757,921 | 1.0% |
| 308 | 286,157 | 0.4% |
| 307 | 229,187 | 0.3% |
| 429 | 116,179 | 0.2% |
2,111 responses were 418 I'm a teapot, and 516 were 451 Unavailable For Legal Reasons. lawa asked for 12,532,017 robots.txt files and, because of what they said, left 2,537,890 pages alone. it walked away from 101,895 servers entirely because of 429, 503, or repeated 403s, and 617,508 linked‐to servers turned out not to exist any more.
Where the W3 is from
| network | servers | share |
|---|---|---|
| Cloudflare, Inc. | 1,366,245 | 22.4% |
| Amazon.com, Inc. | 730,039 | 11.9% |
| Google LLC | 402,357 | 6.6% |
| Squarespace, Inc. | 220,243 | 3.6% |
| Fastly, Inc. | 187,478 | 3.1% |
| Hetzner Online GmbH | 171,457 | 2.8% |
| OVH SAS | 144,212 | 2.4% |
| Cloudflare London, LLC | 138,755 | 2.3% |
| IONOS SE | 117,632 | 1.9% |
| Microsoft Corporation | 78,789 | 1.3% |
| country | servers | share |
|---|---|---|
| United States | 2,043,863 | 45.3% |
| Germany | 573,753 | 12.7% |
| Canada | 307,994 | 6.8% |
| Japan | 186,638 | 4.1% |
| France | 180,517 | 4.0% |
| United Kingdom | 149,315 | 3.3% |
| The Netherlands | 140,989 | 3.1% |
| Switzerland | 88,907 | 2.0% |
| Italy | 61,358 | 1.4% |
| Russia | 60,670 | 1.3% |
Countries leaves out the 1,692,478 servers behind Cloudflare and Fastly, because an anycast address is in every country at once. IP geolocation by DB-IP.
| top‐level domain | servers | share |
|---|---|---|
| .com | 2,832,848 | 45.7% |
| .org | 587,274 | 9.5% |
| .de | 349,215 | 5.6% |
| .net | 194,705 | 3.1% |
| .uk | 171,114 | 2.8% |
| .jp | 131,395 | 2.1% |
| .io | 129,218 | 2.1% |
| .ch | 98,834 | 1.6% |
| .fr | 95,688 | 1.5% |
| .nl | 78,895 | 1.3% |
Non‐standard headers
| header | servers | share |
|---|---|---|
| cf-ray | 1,627,683 | 26.2% |
| cf-cache-status | 1,151,756 | 18.6% |
| x-cache | 521,988 | 8.4% |
| x-served-by | 345,398 | 5.6% |
| x-request-id | 253,394 | 4.1% |
| cf-edge-cache | 227,119 | 3.7% |
| x-proxy-cache | 223,184 | 3.6% |
| x-contextid | 222,445 | 3.6% |
| x-permitted-cross-domain-policies | 220,124 | 3.5% |
| x-cache-hits | 210,318 | 3.4% |
| x-redirect-by | 209,162 | 3.4% |
| x-timer | 196,820 | 3.2% |
1,422 servers (0.02%) send an Onion-Location header, which invites Tor Browser users to a .onion mirror of the site.
And of course, X-Clacks-Overhead. 10,019 servers send it:
| X-Clacks-Overhead | servers | share |
|---|---|---|
| GNU Terry Pratchett | 8,721 | 87.0% |
| GNU Natalie Nguyen | 509 | 5.1% |
| GNU Terry Pratchett, Stephen Hawking, Pr | 46 | 0.5% |
| GNU Terry Pratchett Going Postal | 21 | 0.2% |
| GNU Terry Pratchett, Douglas Adams, Step | 17 | 0.2% |
record holders
| record | value | server |
|---|---|---|
| slowest clock | 56.8 years behind | www.ara.am |
| fastest clock | 136.0 years ahead | edventurebuilder.com |
| oldest Last-Modified | 1991-10-16 | quimby.gnus.org |
| biggest header block | 66,688 bytes | www.mydreamstable.com |
| most header lines | 918 | www.mydreamstable.com |
| most cookies in one response | 910 | www.mydreamstable.com |
| farthest from this site | 42 links away | elliottlevin.com |
35,021 pages claimed implausibly old Last-Modified (before the W3 existed, mostly 1970 (epoch fail!)) and 18,284 claimed one from the future; neither counts.
Details
lawa obeys robots.txt (its name there is “lawa”), waits at least ten seconds between requests to the same server, and takes at most 25 pages from any one. If it is bothering you, a robots.txt rule is honored within a day, and I read my email robin@dreamstation.systems regularly.
Figures cover 2026-09-20 05:43 UTC to 2026-11-05 09:31 UTC (515 data segments); the crawl is constantly running, and the newest ~45 minutes are not in yet.