ROADMAP.md: extended the HTTP/2 section, reformatted

Elaborated on several of the remaining HTTP/2 parts and made document
use a format that ends up nicer on the web page:
http://curl.haxx.se/dev/roadmap.html

docs/ROADMAP.md

@@ -5,44 +5,118 @@ Roadmap of things Daniel Stenberg and Steve Holme want to work on next. It is
 intended to serve as a guideline for others for information, feedback and
 possible participation.
 
-New stuff - libcurl
--------------------
+HTTP/2
+------
+
+- test suite
+
+   Base this on existing nghttp2 server to start with to make functional
+   tests. Later on we can adopt that code or work with nghttp2 to provide ways
+   to have the http2 server respond with broken responses to make sure we deal
+   with that nicely as well.
+
+   To decide: if we need to bundle parts of the nghttp2 stuff that probably
+   won't be shipped by many distros.
+
+- multiplexing/pipelining
+
+   Add a bit to `CURLMOPT_PIPELINING` which then makes libcurl attempt to
+   multiplex HTTP/2 connections to the same host, using the HTTP1 pipelining
+   host matching logic. Doing multiple transfers to the same host using the
+   multi interface will then make them multiplexed. By default, this will
+   handle flow-control for the streams in an automatic fashion.
+
+   With multiplexed streams, we also need to make sure that we terminate
+   streams correctly when a single easy handle is removed from such a
+   connection.
+
+   Since HTTP/2 is designed to always use a single connection for each host
+   (origin really), it seems fair that with "pipelining" enabled there will be
+   no extra checks for number of streams etc added to the connection apart
+   from the maximum number set over the HTTP/2 protocol (the remote sets a
+   maximum) and only if that is reached, another connection gets created.
+
+- stream properties API
+
+   Provide options for setting priorities and dependencies among the streams
+   (easy handles). They are mostly information set for the stream and sent to
+   the server so we don't have to add much logic for this.
+
+- server push
+
+   Not exactly clear exactly how to support this API-wise, but by adding
+   handles without asking for a resource it could be a way to be prepared to
+   receive pushes in case such are sent. We probably need it to still specify
+   a URL with host name, port etc but we probably need a special option to
+   tell libcurl it is for server push purposes.
 
-1. HTTP/2
+- provide option for HTTP/2 "prior knowledge" over clear text
 
- - test suite
- - http2 multiplexing/pipelining
- - provide option for HTTP/2 "prior knowledge" over clear text
- - provide option to allow curl to default to HTTP/2 only when using HTTPS
+   As it would avoid the roundtrip-heavy Upgrade: procedures when you _know_
+   it speaks HTTP/2.
 
-2. SRV records
+- provide option to allow curl to default to HTTP/2 only when using HTTPS
 
-3. HTTPS to proxy
+   We could switch on HTTP/2 by-default for HTTPS quite easily and it
+   shouldn't hurt anyone, while HTTP/2 for HTTP by default could introduce
+   lots of Upgrade: roundtrips that users won't like. So a separated option
+   alternative makes sense.
 
-4. make sure there's an easy handle passed in to `curl_formadd()`,
-   `curl_formget()` and `curl_formfree()` by adding replacement functions and
-   deprecating the old ones to allow custom mallocs and more
+SRV records
+-----------
 
-5. add support for third-party SASL libraries such as Cyrus SASL - may need to
-   move existing native and SSPI based authentication into vsasl folder after
-   reworking HTTP and SASL code
+How to find services for specific domains/hosts.
 
-6. SASL authentication in LDAP
+HTTPS to proxy
+--------------
 
-7. Simplify the SMTP email interface so that programmers don't have to
-   construct the body of an email that contains all the headers, alternative
-   content, images and attachments - maintain raw interface so that
-   programmers that want to do this can
+To avoid network traffic to/from the proxy getting snooped on.
 
-8. Allow the email protocols to return the capabilities before
-    authenticating. This will allow an application to decide on the best
-    authentication mechanism
+curl_formadd()
+--------------
 
-9. Allow Windows threading model to be replaced by Win32 pthreads port
+make sure there's an easy handle passed in to `curl_formadd()`,
+`curl_formget()` and `curl_formfree()` by adding replacement functions and
+deprecating the old ones to allow custom mallocs and more
+
+third-party SASL
+----------------
+
+add support for third-party SASL libraries such as Cyrus SASL - may need to
+move existing native and SSPI based authentication into vsasl folder after
+reworking HTTP and SASL code
+
+SASL authentication in LDAP
+---------------------------
+
+...
+
+Simplify the SMTP email
+-----------------------
+
+Simplify the SMTP email interface so that programmers don't have to
+construct the body of an email that contains all the headers, alternative
+content, images and attachments - maintain raw interface so that
+programmers that want to do this can
+
+email capabilities
+------------------
+
+Allow the email protocols to return the capabilities before
+authenticating. This will allow an application to decide on the best
+authentication mechanism
+
+Win32 pthreads
+--------------
+
+Allow Windows threading model to be replaced by Win32 pthreads port
+
+dynamic buffer size
+-------------------
 
-10. Implement a dynamic buffer size to allow SFTP to use much larger buffers
-    and possibly allow the size to be customizable by applications. Use less
-    memory when handles are not in use?
+Implement a dynamic buffer size to allow SFTP to use much larger buffers and
+possibly allow the size to be customizable by applications. Use less memory
+when handles are not in use?
 
 New stuff - curl
 ----------------