1.  HTTP Fetching in 15 Minutes

    Let's start with a trivial working example.  You'll need a throwaway
    build of the viewer.  And we'll use indra/newview/llappviewer.cpp as
    the host module for these hacks.

    First, add some headers:


        #include "httpcommon.h"
        #include "httprequest.h"
        #include "httphandler.h"


    You'll need to derive a class from HttpHandler (not HttpHandle).
    This is used to deliver notifications of HTTP completion to your
    code.  Place it near the top, before LLDeferredTaskList, say:


        class MyHandler : public LLCore::HttpHandler
        {
        public:
            MyHandler()
            : LLCore::HttpHandler()
            {}

            virtual void onCompleted(LLCore::HttpHandle /* handle */,
                                     LLCore::HttpResponse * /* response */)
            {
                LL_INFOS("Hack") << "It is happening again." << LL_ENDL;

                delete this;    // Last statement
            }
        };


    Add some statics up there as well:


        // Our request object.  Allocate during initialiation.
        static LLCore::HttpRequest * my_request(NULL);

        // The policy class for HTTP traffic.
        // Use HttpRequest::DEFAULT_POLICY_ID, but DO NOT SHIP WITH THIS VALUE!!
        static LLCore::HttpRequest::policy_t my_policy(LLCore::HttpRequest::DEFAULT_POLICY_ID);

        // Priority for HTTP requests.  Use 0U.
        static LLCore::HttpRequest::priority_t my_priority(0U);


    In LLAppViewer::init() after mAppCoreHttp.init(), create a request object:


        my_request = new LLCore::HttpRequest();


    In LLAppViewer::mainLoop(), just before entering the while loop,
    we'll kick off one HTTP request:


        // Construct a handler object (we'll use the heap this time):
        MyHandler * my_handler = new MyHandler;

        // Issue a GET request to 'http://www.example.com/' kicking off
        // all the I/O, retry logic, etc.
        LLCore::HttpHandle handle;
        handle = my_request->requestGet(my_policy,
                                        my_priority,
                                        "http://www.example.com/",
                                        NULL,
                                        NULL,
                                        my_handler);
        if (LLCORE_HTTP_HANDLE_INVALID == handle)
        {
            LL_WARNS("Hack") << "Failed to launch HTTP request.  Try again."
                             << LL_ENDL;
        }


    Finally, arrange to periodically call update() on the request object
    to find out when the request completes.  This will be done by
    calling the onCompleted() method with status information and
    response data from the HTTP operation.  Add this to the
    LLAppViewer::idle() method after the ping:


        my_request->update(0);


    That's it.  Build it, run it and watch the log file.  You should get
    the "It is happening again." message indicating that the HTTP
    operation completed in some manner.


2.  What Does All That Mean

    MyHandler/HttpHandler.  This class replaces the Responder-style in
    legacy code.  One method is currently defined.  It is used for all
    request completions, successful or failed:


        void onCompleted(LLCore::HttpHandle /* handle */,
                         LLCore::HttpResponse * /* response */);


    The onCompleted() method is invoked as a callback during calls to
    HttpRequest::update().  All I/O is completed asynchronously in
    another thread.  But notifications are polled by calling update()
    and invoking a handler for completed requests.

    In this example, the invocation also deletes the handler (which is
    never referenced by the llcorehttp code again).  But other
    allocation models are possible including handlers shared by many
    requests, stack-based handlers and handlers mixed in with other,
    unrelated classes.

    LLCore::HttpRequest().  Instances of this class are used to request
    all major functions of the library.  Initialization, starting
    requests, delivering final notification of completion and various
    utility operations are all done via instances.  There is one very
    important rule for instances:

        Request objects may NOT be shared between threads.

    my_priority.  The APIs support the idea of priority ordering of
    requests but it hasn't been implemented and the hope is that this
    will become useless and removed from the interface.  Use 0U except
    as noted.

    my_policy.  This is an important one.  This library attempts to
    manage TCP connection usage more rigorously than in the past.  This
    is done by issuing requests to a queue that has various settable
    properties.  These establish connection usage for the queue as well
    as how queues compete with one another.  (This is patterned after
    class-based queueing used in various networking stacks.)  Several
    classes are pre-defined.  Deciding when to use an existing class and
    when to create a new one will determine what kind of experience
    users have.  We'll pick up this question in detail below.

    requestGet().  Issues an ordinary HTTP GET request to a given URL
    and associating the request with a policy class, a priority and an
    response handler.  Two additional arguments, not used here, allow
    for additional headers on the request and for per-request options.
    If successful, the call returns a handle whose value is other than
    LLCORE_HTTP_HANDLE_INVALID.  The HTTP operation is then performed
    asynchronously by another thread without any additional work by the
    caller.  If the handle returned is invalid, you can get the status
    code by calling my_request->getStatus().

    update().  To get notification that the request has completed, a
    call to update() will invoke onCompleted() methods.


3.  Refinements, Necessary and Otherwise

    MyHandler::onCompleted().  You'll want to do something useful with
    your response.  Distinguish errors from successes and getting the
    response body back in some form.

    Add a new header:


        #include "bufferarray.h"


    Replace the existing MyHandler::onCompleted() definition with:


        virtual void onCompleted(LLCore::HttpHandle /* handle */,
                                 LLCore::HttpResponse * response)
        {
            LLCore::HttpStatus status = response->getStatus();
            if (status)
            {
                // Successful request.  Try to fetch the data
                LLCore::BufferArray * data = response->getBody();

                if (data && data->size())
                {
                    // There's some data.  A BufferArray is a linked list
                    // of buckets.  We'll create a linear buffer and copy
                    // the data into it.
                    size_t data_len = data->size();
                    char * data_blob = new char [data_len + 1];
                    data->read(0, data_blob, data_len);
                    data_blob[data_len] = '\0';

                    // Process the data now in NUL-terminated string.
                    // Needs more scrubbing but this will do.
                    LL_INFOS("Hack") << "Received:  " << data_blob << LL_ENDL;

                    // Free the temporary data
                    delete [] data_blob;
                }
            }
            else
            {
                // Something went wrong.  Translate the status to
                // a meaningful message.
                LL_WARNS("Hack") << "HTTP GET failed.  Status:  "
                                 << status.toTerseString()
                                 << ", Reason:  " << status.toString()
                                 << LL_ENDL;
            }           

            delete this;    // Last statement
        }


    HttpHeaders.  The header file "httprequest.h" documents the expected
    important headers that will go out with the request.  You can add to
    these by including an HttpHeaders object with the requestGet() call.
    These are typically setup once as part of init rather than
    dynamically created.

    Add another header:


        #include "httpheaders.h"


    In LLAppViewer::mainLoop(), add this alongside the allocation of
    my_handler:


        // Additional headers for all requests
        LLCore::HttpHeaders * my_headers = new LLCore::HttpHeaders();
        my_headers->append("Accept", "text/html, application/llsd+xml");


    HttpOptions.  Options are similar and include a mix of value types.
    One interesting per-request option is the trace setting.  This
    enables various debug-type messages in the log file that show the
    progress of the request through the library.  It takes values from
    zero to three with higher values giving more verbose logging.  We'll
    use '2' and this will also give us a chance to verify that
    HttpHeaders works as expected.

    Same as above, a new header:


        #include "httpoptions.h"


    And in LLAppView::mainLoop():


        // Special options for requests
        LLCore::HttpOptions * my_options = new LLCore::HttpOptions();
        my_options->setTrace(2);


    Now let's put that all together into a more complete requesting
    sequence.  Replace the existing invocation of requestGet() with this
    slightly more elaborate block:


        LLCore::HttpHandle handle;
        handle = my_request->requestGet(my_policy,
                                        my_priority,
                                        "http://www.example.com/",
                                        my_options,
                                        my_headers,
                                        my_handler);
        if (LLCORE_HTTP_HANDLE_INVALID == handle)
        {
             LLCore::HttpStatus status = my_request->getStatus();

             LL_WARNS("Hack") << "Failed to request HTTP GET.  Status:  "
                              << status.toTerseString()
                              << ", Reason:  " << status.toString()
                              << LL_ENDL;

             delete my_handler;    // No longer needed.
             my_handler = NULL;
        }


    Build, run and examine the log file.  You'll get some new data with
    this run.  First, you should get the www.example.com home page
    content:


----------------------------------------------------------------------------
2013-09-17T20:26:51Z INFO: MyHandler::onCompleted: Received:  <!doctype html>
<html>
<head>
    <title>Example Domain</title>

    <meta charset="utf-8" />
    <meta http-equiv="Content-type" content="text/html; charset=utf-8" />
    <meta name="viewport" content="width=device-width, initial-scale=1" />
    <style type="text/css">
    body {
        background-color: #f0f0f2;
        margin: 0;
        padding: 0;
        font-family: "Open Sans", "Helvetica Neue", Helvetica, Arial, sans-serif;
        
    }
    div {
        width: 600px;
        margin: 5em auto;
        padding: 50px;
        background-color: #fff;
        border-radius: 1em;
    }
    a:link, a:visited {
        color: #38488f;
        text-decoration: none;
    }
    @media (max-width: 700px) {
        body {
            background-color: #fff;
        }
        div {
            width: auto;
            margin: 0 auto;
            border-radius: 0;
            padding: 1em;
        }
    }
    </style>    
</head>

<body>
<div>
    <h1>Example Domain</h1>
    <p>This domain is established to be used for illustrative examples in documents. You may use this
    domain in examples without prior coordination or asking for permission.</p>
    <p><a href="http://www.iana.org/domains/example">More information...</a></p>
</div>
</body>
</html>
----------------------------------------------------------------------------


    You'll also get a detailed trace of the HTTP operation itself.  Note
    the HEADEROUT line which shows the additional header added to the
    request.


----------------------------------------------------------------------------
HttpService::processRequestQueue: TRACE, FromRequestQueue, Handle:  086D3148
HttpLibcurl::addOp: TRACE, ToActiveQueue, Handle:  086D3148, Actives:  0, Readies:  0
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  TEXT, Data:  About to connect() to www.example.com port 80 (#0) 
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  TEXT, Data:    Trying 93.184.216.119... 
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  TEXT, Data:  Connected to www.example.com (93.184.216.119) port 80 (#0) 
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  TEXT, Data:  Connected to www.example.com (93.184.216.119) port 80 (#0) 
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  HEADEROUT, Data:  GET / HTTP/1.1  Host: www.example.com  Accept-Encoding: deflate, gzip  Connection: keep-alive  Keep-alive: 300  Accept: text/html, application/llsd+xml
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  HEADERIN, Data:  HTTP/1.1 200 OK  
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  HEADERIN, Data:  Accept-Ranges: bytes  
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  HEADERIN, Data:  Cache-Control: max-age=604800  
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  HEADERIN, Data:  Content-Type: text/html  
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  HEADERIN, Data:  Date: Tue, 17 Sep 2013 20:26:56 GMT  
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  HEADERIN, Data:  Etag: "3012602696"  
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  HEADERIN, Data:  Expires: Tue, 24 Sep 2013 20:26:56 GMT  
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  HEADERIN, Data:  Last-Modified: Fri, 09 Aug 2013 23:54:35 GMT  
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  HEADERIN, Data:  Server: ECS (ewr/1590)  
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  HEADERIN, Data:  X-Cache: HIT  
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  HEADERIN, Data:  x-ec-custom-error: 1  
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  HEADERIN, Data:  Content-Length: 1270  
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  HEADERIN, Data:    
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  DATAIN, Data:  256 Bytes
HttpOpRequest::debugCallback: TRACE, LibcurlDebug, Handle:  086D3148, Type:  TEXT, Data:  Connection #0 to host www.example.com left intact 
HttpLibcurl::completeRequest: TRACE, RequestComplete, Handle:  086D3148, Status:  Http_200
HttpOperation::addAsReply: TRACE, ToReplyQueue, Handle:  086D3148
----------------------------------------------------------------------------


4.  What Does All That Mean, Part 2

    HttpStatus.  The HttpStatus object encodes errors from libcurl, the
    library itself and HTTP status values.  It does this to avoid
    collapsing all non-HTTP error into a single '499' HTTP status and to
    make errors distinct.

    To aid programming, the usual bool conversions are available so that
    you can write 'if (status)' and the expected thing will happen
    whether it's an HTTP, libcurl or library error.  There's also
    provision to override the treatment of HTTP errors (making 404 a
    success, say).

    Share data, don't copy it.  The library was started with the goal of
    avoiding data copies as much as possible.  Instead, read-only data
    sharing across threads with atomic reference counts is used for a
    number of data types.  These currently are:

        * BufferArray.  Linked list of data blocks/HTTP bodies.
        * HttpHeaders.  Shared headers for both requests and responses.
        * HttpOptions.  Request-only data modifying HTTP behavior.
        * HttpResponse.  HTTP response description given to onCompleted.

    Using objects of these types requires a few rules:

        * Constructor always gives a reference to caller.
        * References are dropped with release() not delete.
        * Additional references may be taken out with addRef().
        * Unless otherwise stated, once an object is shared with another
          thread it should be treated as read-only.  There's no
          synchronization on the objects themselves.

    HttpResponse.  You'll encounter this mainly in onCompleted() methods.
    Commonly-used interfaces on this object:

        * getStatus() to return the final status of the request.
        * getBody() to retrieve the response body which may be NULL or
          zero-length.
        * getContentType() to return the value of the 'Content-Type'
          header or an empty string if none was sent.

    This is a reference-counted object so you can call addRef() on it
    and hold onto the response for an arbitrary time.  But you'll
    usually just call a few methods and return from onCompleted() whose
    caller will release the object.

    BufferArray.  The core data representation for request and response
    bodies.  In HTTP responses, it's fetched with the getBody() method
    and may be NULL or non-NULL with zero length.  All successful data
    handling should check both conditions before attempting to fetch
    data from the object.  Data access model uses simple read/write
    semantics:

        * append()
        * size()
        * read()
        * write()

    (There is a more sophisticated stream adapter that extends these
    methods and will be covered below.)  So, one way to retrieve data
    from a request is as follows:


        LLCore::BufferArray * data = response->getBody();
        if (data && data->size())
        {
            size_t data_len = data->size();
            char * data_blob = new char [data_len + 1];
            data->read(0, data_blob, data_len);


    HttpOptions and HttpResponse.  Really just simple containers of POD
    and std::string pairs.  But reference counted and the rule about not
    modifying after sharing must be followed.  You'll have the urge to
    change options dynamically at some point.  And you'll try to do that
    by just writing new values to the shared object.  And in tests
    everything will appear to work.  Then you ship and people in the
    real world start hitting read/write races in strings and then crash.
    Don't be lazy.

    HttpHandle.  Uniquely identifies a request and can be used to
    identify it in an onCompleted() method or cancel it if it's still
    queued.  But as soon as a request's onCompleted() invocation
    returns, the handle becomes invalid and may be reused immediately
    for new requests.  Don't hold on to handles after notification.


5.  And Still More Refinements

    (Note: The following refinements are just code fragments.  They
    don't directly fit into the working example above.  But they
    demonstrate several idioms you'll want to copy.)

    LLSD, std::streambuf, std::iostream.  The read(), write() and
    append() methods may be adequate for your purposes.  But we use a
    lot of LLSD.  Its interfaces aren't particularly compatible with
    BufferArray.  And so two adapters are available to give
    stream-like behaviors:  BufferArrayStreamBuf and BufferArrayStream,
    which implement the std::streambuf and std::iostream interfaces,
    respectively.

    A std::streambuf interface isn't something you'll want to use
    directly.  Instead, you'll use the much friendlier std::iostream
    interface found in BufferArrayStream.  This adapter gives you all
    the '>>' and '<<' operators you'll want as well as working
    directly with the LLSD conversion operators.

    Some new headers:


        #include "bufferstream.h"
        #include "llsdserialize.h"


    And an updated fragment based on onCompleted() above:


                // Successful request.  Try to fetch the data
                LLCore::BufferArray * data = response->getBody();
                LLSD resp_llsd;

                if (data && data->size())
                {
                    // There's some data and we expect this to be
                    // LLSD.  Checking of content type and validation
                    // during parsing would be admirable additions.
                    // But we'll forgo that now.
                    LLCore::BufferArrayStream data_stream(data);
                    LLSDSerialize::fromXML(resp_llsd, data_stream);
                }
                LL_INFOS("Hack") << "LLSD Received:  " << resp_llsd << LL_ENDL;
            }
            else
            {


    Converting an LLSD object into an XML stream stored in a
    BufferArray is just the reverse of the above:


        BufferArray * data = new BufferArray();
        LLCore::BufferArrayStream data_stream(data);

        LLSD src_llsd;
        src_llsd["foo"] = "bar";

        LLSDSerialize::toXML(src_llsd, data_stream);

        // 'data' now contains an XML payload and can be sent
        // to a web service using the requestPut() or requestPost()
        //  methods.
        ...  requestPost(...);

        // And don't forget to release the BufferArray.
        data->release();
        data = NULL;


    There are now helper functions in llmessage/llcorehttputil.h to
    assist with LLSD usage.  requestPostWithLLSD(...) provides a
    requestPost()-like interface that takes an LLSD object rather than
    a BufferArray.  And responseToLLSD(...) attempts to convert a
    BufferArray received from a server into an LLSD object.  You can
    find examples in llmeshrepository.cpp, llinventorymodel.cpp,
    llinventorymodelbackgroundfetch.cpp and lltexturefetch.cpp.

    LLSD will often go hand-in-hand with BufferArray and data
    transport.  But you can also do all the streaming I/O you'd expect
    of a std::iostream object:


        BufferArray * data = new BufferArray();
        LLCore::BufferArrayStream data_stream(data);

        data_stream << "Hello, World!" << 29.4 << '\n';
        std::string str;
        data_stream >> str;
        std::cout << str << std::endl;

        data->release();
        // Actual delete will occur when 'data_stream'
        // falls out of scope and is destructed.


    Scoping objects and cleaning up.  The examples haven't bothered
    with cleanup of objects that are no longer needed.  Instead, most
    objects have been allocated as if they were global and eternal.
    You'll put the objects in more appropriate feature objects and
    clean them up as a group.  Here's a checklist for actions you may
    need to take on cleanup:

    * Call delete on:
      o HttpHandlers created on the heap
      o HttpRequest objects
    * Call release() on:
      o BufferArray objects
      o HttpHeaders objects
      o HttpOptions objects
      o HttpResponse objects

    On program exit, as threads wind down, the library continues to
    operate safely.  Threads don't interact via the library and even
    dangling references to HttpHandler objects are safe.  If you don't
    call HttpRequest::update(), handler references are never
    dereferenced.

    You can take a more thorough approach to wind-down.  Keep a list
    of HttpHandles (not HttpHandlers) of outstanding requests.  For
    each of these, call HttpRequest::requestCancel() to cancel the
    operation.  (Don't add the cancel requests' handled to the list.)
    This will cancel the outstanding requests that haven't completed.
    Canceled or completed, all requests will queue notifications.  You
    can now cycle calling update() discarding responses.  Continue
    until all requests notify or a few seconds have passed.

    Global startup and shutdown is handled in the viewer.  But you can
    learn about it in the code or in the documentation in the headers.


6.  Choosing a Policy Class

    Now it's time to get rid of the default policy class.  Take a look
    at the policy class definitions in newview/llappcorehttp.h.
    Ideally, you'll find one that's compatible with what you're doing.
    Some of the compatibility guidelines are:

    * Destination: Pair of host and port.  Mixing requests with
      different destinations may cause more connection setup and tear
      down.

    * Method: http or https.  Usually moot given destination.  But
      mixing these may also cause connection churn.

    * Transfer size: If you're moving 100MB at a time and you make your
      requests to the same policy class as a lot of small, fast event
      information that fast traffic is going to get stuck behind you
      and someone's experience is going to be miserable.

    * Long poll requests: These are long-lived, must- do operations.
      They have a special home called AP_LONG_POLL.

    * Concurrency: High concurrency (5 or more) and large transfer
      sizes are incompatible.  Another head-of-the-line problem.  High
      concurrency is tolerated when it's desired to get maximal
      throughput.  Mesh and texture downloads, for example.

    * Pipelined: If your requests are not idempotent, stay away from
      anything marked 'soon' or 'yes'.  Hidden retries may be a
      problem for you.  For now, would also recommend keeping PUT and
      POST requests out of classes that may be pipelined.  Support for
      that is still a bit new.

    If you haven't found a compatible match, you can either create a
    new class (llappcorehttp.*) or just use AP_DEFAULT, the catchall
    class when all else fails.  Inventory query operations might be a
    candidate for a new class that supported pipelining on https:.
    Same with display name lookups and other bursty-at-login
    operations.  For other things, AP_DEFAULT will do what it can and
    will, in some way or another, tolerate any usage.  Whether the
    users' experiences are good are for you to determine.

    
7.  FAQ

    Q1.  What do these policy classes achieve?

    A1.  Previously, HTTP-using code in the viewer was written as if
    it were some isolated, local operation that didn't have to
    consider resources, contention or impact on services and the
    larger environment.  The result was an application with on the
    order of 100 HTTP launch points in its codebase that could create
    dozens or even 100's of TCP connections zeroing in on grid
    services and disrupting networking equipment, web services and
    innocent users.  The use of policy classes (modeled on
    http://en.wikipedia.org/wiki/Class-based_queueing) is a means to
    restrict connection concurrency, good and necessary in itself.  In
    turn, that reduces demands on an expensive resource (connection
    setup and concurrency) which relieves strain on network points.
    That enables connection keepalive and opportunites for true
    improvements in throughput and user experience.

    Another aspect of the classes is that they give some control over
    how competing demands for the network will be apportioned.  If
    mesh fetches, texture fetches and inventory queries are all being
    made at once, the relative weights of their classes' concurrency
    limits established that apportioning.  We now have an opportunity
    to balance the entire viewer system.

    Q2.  How's that data sharing with refcounts working for you?

    A2.  Meh.  It does reduce memory churn and the frequency at which
    free blocks must be moved between threads.  But it's also a design
    for static configuration and dynamic reconfiguration (not
    requiring a restart) is favored.  Creating new options for every
    request isn't too bad, it a sequence of "new, fill, request,
    release" for each requested operation.  That in contrast to doing
    the "new, fill, release" at startup.  The bad comes in getting at
    the source data.  One rule in this work was "no new thread
    problems."  And one source for those is pulling setting values out
    of gSettings in threads.  None of that is thread safe though we
    tend to get away with it.

    Q3.  What needs to be done?

    A3.  There's a To-Do list in _httpinternal.h.  It has both large
    and small projects here if someone would like to try changes.