- Removes the __fetch namespace from `deno types`
- Response.redirect should be a static.
- Response.body should not be AsyncIterable.
- Disables the deno_proxy benchmark
- Makes std/examples/curl.ts buffer the body before printing to stdout
I just tried it and found that using application/typescript, the browser will download
the file directly, I think that .ts should be mapped to application/javascript or
text/typescript
This relates directly to [an
issue](https://github.com/denoland/deno_std/issues/620) that I initially
raised in `deno_std` awhile back, and was reminded about it today when
the `oak` project popped up on my github recommended repos.
As of now Deno's http servers are vulnerable to the same underlying
issue of go CVE-2019-16276 due to the fact that it's based off of ported
go code from their old standard library. [Here's the commit that fixed
the
CVE.](6e6f4aaf70)
Long story short, some off the shelf proxies and caching servers allow
for passing unaltered malformed headers to backends that they're
fronting. When they pass invalid headers that they don't understand this
can cause issues with HTTP request smuggling. I believe that to this
date, this is the default behavior of AWS ALBs--meaning any server that
strips whitespace from the tail end of header field names and then
interprets the header, when placed behind an ALB, is susceptible to
request smuggling.
The current behavior is actually specifically called out in [RFC
7230](https://tools.ietf.org/html/rfc7230#section-3.2.4) as something
that MUST result in a rejected message, but the change corresponding to
this PR, is more lenient and what both go and nginx currently do, and is
better than the current behavior.
* Reduce "testing" interfaces
* Use a callback instead of a generator for Deno.runTests()
* Default RunTestsOptions::reportToConsole to true
* Compose TestMessage into a single interface
* Remove DENO_BUILD_MODE and DENO_BUILD_PATH
Also remove outdated docs related to ninja/gn.
* fix
* remove parameter to build_mode()
* remove arg parsing from benchmark.py
Due to structure of "Server" for each open connection there's a pending "read" op. Because connection owned by "Server" are not tracked, calling "Server.close()" doesn't close open connections.
This commit introduces simple tracking of connections for server and ensures owned connections are closed on "Server.close()".
