--- a/taskcluster/ci/source-test/doc.yml
+++ b/taskcluster/ci/source-test/doc.yml
@@ -7,23 +7,22 @@ doc-generate:
         tier: 1
     worker-type: aws-provisioner-v1/gecko-t-linux-xlarge
     worker:
         docker-image: {in-tree: "lint"}
         max-run-time: 1800
         artifacts:
             - type: file
               name: public/docs.tar.gz
-              path: /home/worker/checkouts/gecko/docs.tar.gz
+              path: /home/worker/checkouts/gecko/docs-out/main.tar.gz
     run:
         using: run-task
         command: >
             cd /home/worker/checkouts/gecko &&
-            ./mach doc --outdir docs-out --no-open --archive &&
-            mv docs-out/Mozilla_Source_Tree_Docs.tar.gz docs.tar.gz
+            ./mach doc --outdir docs-out --no-open --archive
     when:
         files-changed:
             - '**/*.py'
             - '**/*.rst'
             - 'tools/docs/**'
 
 doc-upload:
     description: Generate and upload the Sphinx documentation

--- a/tools/docs/conf.py
+++ b/tools/docs/conf.py
@@ -76,8 +76,10 @@ else:
     # environment handles this otherwise.
     import sphinx_rtd_theme
     html_theme = 'sphinx_rtd_theme'
     html_theme_path = [sphinx_rtd_theme.get_html_theme_path()]
 
 
 html_static_path = ['_static']
 htmlhelp_basename = 'MozillaTreeDocs'
+
+moz_project_name = 'main'

--- a/tools/docs/mach_commands.py
+++ b/tools/docs/mach_commands.py
@@ -61,40 +61,39 @@ class Documentation(MachCommandBase):
         for path in what:
             path = os.path.normpath(os.path.abspath(path))
             docdir = self._find_doc_dir(path)
 
             if not docdir:
                 failed.append((path, 'could not find docs at this location'))
                 continue
 
-            # find project name to use as a namespace within `outdir`
-            project = self._find_project_name(docdir)
-            savedir = os.path.join(format_outdir, project)
+            props = self._project_properties(docdir)
+            savedir = os.path.join(format_outdir, props['project'])
 
             args = [
                 'sphinx',
                 '-b', format,
                 docdir,
                 savedir,
             ]
             result = sphinx.build_main(args)
             if result != 0:
                 failed.append((path, 'sphinx return code %d' % result))
             else:
                 generated.append(savedir)
 
             if archive:
                 archive_path = os.path.join(outdir,
-                                            '%s.tar.gz' %  project)
+                                            '%s.tar.gz' % props['project'])
                 moztreedocs.create_tarball(archive_path, savedir)
                 print('Archived to %s' % archive_path)
 
             if upload:
-                self._s3_upload(savedir)
+                self._s3_upload(savedir, props['project'], props['version'])
 
             index_path = os.path.join(savedir, 'index.html')
             if not http and auto_open and os.path.isfile(index_path):
                 webbrowser.open(index_path)
 
         if generated:
             print('\nGenerated documentation:\n%s\n' % '\n'.join(generated))
 
@@ -108,37 +107,63 @@ class Documentation(MachCommandBase):
             if len(addr) != 2:
                 return die('invalid address: %s' % http)
 
             httpd = mozhttpd.MozHttpd(host=addr[0], port=addr[1],
                                       docroot=format_outdir)
             print('listening on %s:%d' % addr)
             httpd.start(block=True)
 
-    def _find_project_name(self, path):
+    def _project_properties(self, path):
         import imp
         path = os.path.join(path, 'conf.py')
         with open(path, 'r') as fh:
             conf = imp.load_module('doc_conf', fh, path,
                                    ('.py', 'r', imp.PY_SOURCE))
 
-        return conf.project.replace(' ', '_')
+        # Prefer the Mozilla project name, falling back to Sphinx's
+        # default variable if it isn't defined.
+        project = getattr(conf, 'moz_project_name', None)
+        if not project:
+            project = conf.project.replace(' ', '_')
+
+        return {
+            'project': project,
+            'version': getattr(conf, 'version', None)
+        }
 
     def _find_doc_dir(self, path):
         search_dirs = ('doc', 'docs')
         for d in search_dirs:
             p = os.path.join(path, d)
             if os.path.isfile(os.path.join(p, 'conf.py')):
                 return p
 
-    def _s3_upload(self, root):
+    def _s3_upload(self, root, project, version=None):
         self.virtualenv_manager.install_pip_package('boto3==1.4.4')
 
         from moztreedocs import distribution_files
         from moztreedocs.upload import s3_upload
-        files = distribution_files(root)
-        s3_upload(files)
+
+        # Files are uploaded to multiple locations:
+        #
+        # <project>/latest
+        # <project>/<version>
+        #
+        # This allows multiple projects and versions to be stored in the
+        # S3 bucket.
+
+        files = list(distribution_files(root))
+
+        s3_upload(files, key_prefix='%s/latest' % project)
+        if version:
+            s3_upload(files, key_prefix='%s/%s' % (project, version))
+
+        # Until we redirect / to main/latest, upload the main docs
+        # to the root.
+        if project == 'main':
+            s3_upload(files)
 
 
 def die(msg, exit_code=1):
     msg = '%s: %s' % (sys.argv[0], msg)
     print(msg, file=sys.stderr)
     return exit_code

--- a/tools/docs/moztreedocs/upload.py
+++ b/tools/docs/moztreedocs/upload.py
@@ -7,17 +7,25 @@ from __future__ import absolute_import, 
 import io
 import mimetypes
 import os
 
 import boto3
 import requests
 
 
-def s3_upload(files):
+def s3_upload(files, key_prefix=None):
+    """Upload files to an S3 bucket.
+
+    ``files`` is an iterable of ``(path, BaseFile)`` (typically from a
+    mozpack Finder).
+
+    Keys in the bucket correspond to source filenames. If ``key_prefix`` is
+    defined, key names will be ``<key_prefix>/<path>``.
+    """
     region = 'us-west-2'
     level = os.environ.get('MOZ_SCM_LEVEL', '1')
     bucket = {
         '1': 'gecko-docs.mozilla.org-l1',
         '2': 'gecko-docs.mozilla.org-l2',
         '3': 'gecko-docs.mozilla.org',
     }[level]
     secrets_url = 'http://taskcluster/secrets/v1/secret/'
@@ -43,13 +51,20 @@ def s3_upload(files):
 
     for path, f in files:
         content_type, content_encoding = mimetypes.guess_type(path)
         extra_args = {}
         if content_type:
             extra_args['ContentType'] = content_type
         if content_encoding:
             extra_args['ContentEncoding'] = content_encoding
-        print('uploading', path)
+
+        if key_prefix:
+            key = '%s/%s' % (key_prefix, path)
+        else:
+            key = path
+
+        print('uploading %s to %s' % (path, key))
+
         # The file types returned by mozpack behave like file objects. But they
         # don't accept an argument to read(). So we wrap in a BytesIO.
-        s3.upload_fileobj(io.BytesIO(f.read()), bucket, path,
+        s3.upload_fileobj(io.BytesIO(f.read()), bucket, key,
                           ExtraArgs=extra_args)