versions of CMU CL, as well as guarding against short reads by
read-sequence on socket streams, though my understanding of the
standard seems to imply that read-sequence can only return prematurely
on EOF. But this way we shouldn't have any problems either way.
Added naive buffering functionality for use in CLASH. This is
currently only used to implement stream entities, but might have other
uses. Given the exitence of generational garbage collectors, it might
be advantageous to forego the pooling scheme and throw away buffers
after each use. This will depend on implementations and usage
patterns, though, so we should make this tunable...
Fixed various bugs in query-argument parsing (don't you just love
ambiguous standards?), and added functionality to default url slots
from another url object, to copy url objects and to externalize them.
Code to implement session management for resources. The current
version supports transient cookie based session management, but other
forms of management could be added without undue hassle.
The machinery in this file is used for the automatic creation of
responses in standardized (error) situations based on the information
contained in a condition.
Changed argument precedence order on access-resource and
access-resource-using-method to implement the expected semantics that
methods specialized on requests or request-methods overide more
general methods in base classes.
Also revamped the division of labour between entities and resources on
static resources (see entity.cl), so that static resources are simple
wrappers around static entities, which provide the content.
Added resources that handle dynamic content and forms by dispatching
to functions passed in during creation. This is to support simple
ad-hoc dynamic resources, whereas really dynamic resources will just
subclass resource and implement the stuff on their own.
Added a simple wrapper to bind form data passed in with a post request
(with-form-data). This should be unified with query-argument
processing.
Moved some of the common parsing and rendering subroutines to a new
file. Many other functions should probably be moved over here, so
watch this space as clean up progresses further.
Extensive changes to accomodate the new model for message classes and
HTTP versions. This was necessary because the old model was not as
flexible and concise as hoped, so that supporting multiple http
version became tedious. The new infrastructure is still a bit of a
work in progress, though, so watch this space. There's also some work
left purging the reliance on pure streams instead of connections.
Added simple common logging file format mixin and a simple profiling
mixin. This is more for demonstration purposes, than for real-life
usage. Real world appliations will probably want to do their own
customized forms of logging and/or profiling.
Moved render-slots to messages.cl, removed non-1.0 entity-headers,
and implemented the rest. Added a number of ready-made entity
implementations, for use with static-resource.
Consolidated the event-driven CMU CL driver code and the
multi-processing CMU CL driver so that they share common code, like
the newly introduced connection sub-class and related methods.
Made connection an abstract base class from which individual drivers
deriver their specific connection classes (like it was intended in the
first place). Added generic functions to get the address and hostname
of the client.
Added response-initargs and entity-initargs fields to the clash-error
conditions, so that signallers can return additional information to
the client regarding errors.
all versions of CMUCL that don't provide multi-processing support,
like on all non-x86 platforms. Tested on an UltraSparc running
Solaris 2.7 (hawk) and a mid-1999 version of CMUCL 18b+.
handling, in the face of the much saner RFC 2396. This implementation
provides all major capabilities, including low-consing operation,
representation of relative and absolute URIs, useful merging and
read-write consistency of URIs. Some clean-up work is still needed to
purge the implementation from all references to URLs, to support the
handling of all URIs. Other minor cleanups and additions are still
needed, but after that the handling of URIs should be nearly as robust
as the handling of pathnames in CL.
account the server we are servicing, so that each server only sees
and creates request/response classes he knows he can handle. This
also makes it possible to add request/response classes in a running
system, and to run servers of different versions side-by-side in an
image without affecting each other...
nearly all concepts and simple implementations thereof needed to get a
simple HTTP/0.9 "compliant" server working (there are some hacks
needed that we don't yet provide, since the correct things will be
added shortly, like complete HTTP/1.1 request parsing. The hacks
needed are provided as part of the basic HTTP/0.9 server demo in
src/test/basic-demo.cl).
Further work is needed to clean up some things, Entity and Resource
handling needs to be implemented right and less "naive" (the current
implementations are just simple place-holders to get things up and
running). Connections need to have knowledge of client identity
(passed from the driver, this is implementation-specific stuff).
Logging needs to be implemented (probably as server mixins).
Condition handling needs to generate better responses for HTTP/0.9,
and the division between condition handling and normal handling needs
to be documented/rethought. Content generation is totally missing
currently and needs to be implemented. If this is all in place, an
HTTP/1.0 conforming server should be possible, and after porting the
drivers to ACL and LW, we can make a first release.
