[] (0) Allow UAs to opt out of the storage mutex and corrupt data ins…

…tead of being slow. Fixing http://www.w3.org/Bugs/Public/show_bug.cgi?id=13020 git-svn-id: http://svn.whatwg.org/webapps@6368 340c8d12-0b0e-0410-8428-c7bf67bfef74
whatwg · Aug 4, 2011 · 56cf59c · 56cf59c
1 parent c5c1b19
commit 56cf59c
Show file tree

Hide file tree

Showing 3 changed files with 64 additions and 15 deletions.
diff --git a/complete.html b/complete.html
@@ -67995,11 +67995,27 @@ <h5 id=definitions-1><span class=secno>7.1.4.1 </span>Definitions</h5>
   was queued; if the task was queued by or for a <a href=#concept-script title=concept-script>script</a> then the document is the
   <a href="#script's-document">script's document</a>.</p>
 
-  <p>A user agent is required to have one <dfn id=storage-mutex>storage
-  mutex</dfn>. This mutex is used to control access to shared state
-  like cookies. At any one point, the <a href=#storage-mutex>storage mutex</a> is
-  either free, or owned by a particular <a href=#event-loop>event loop</a> or
-  instance of the <a href=#fetch title=fetch>fetching</a> algorithm.</p>
+  <p>A user agent may have one <dfn id=storage-mutex>storage mutex</dfn>. This mutex is
+  used to control access to shared state like cookies. At any one
+  point, the <a href=#storage-mutex>storage mutex</a> is either free, or owned by a
+  particular <a href=#event-loop>event loop</a> or instance of the <a href=#fetch title=fetch>fetching</a> algorithm.</p>
+
+  <p>If a user agent does not implement a <a href=#storage-mutex>storage mutex</a>,
+  it is exempt from implementing the requirements that require it to
+  acquire or release it.</p>
+
+  <p class=note>User agent implementors have to make a choice
+  between two evils. On the one hand, not implementing the storage
+  mutex means that there is a risk of data corruption: a site could,
+  for instance, try to read a cookie, increment its value, then write
+  it back out, using the new value of the cookie as a unique
+  identifier for the session; if the site does this twice in two
+  different browser windows at the same time, it might end up using
+  the same "unique" identifier for both sessions, with potentially
+  disastrous effects. On the other hand, implementing the storage
+  mutex has potentially serious performance implications: whenever a
+  site uses Web Storage or cookies, all other sites that try to use
+  Web Storage or cookies are blocked until the first site finishes.</p>
 
   <p>Whenever a <a href=#concept-script title=concept-script>script</a> calls into
   a <a href=#plugin>plugin</a>, and whenever a <a href=#plugin>plugin</a> calls into

diff --git a/index b/index
@@ -67887,11 +67887,27 @@ interface <dfn id=navigatoronline>NavigatorOnLine</dfn> {
   was queued; if the task was queued by or for a <a href=#concept-script title=concept-script>script</a> then the document is the
   <a href="#script's-document">script's document</a>.</p>
 
-  <p>A user agent is required to have one <dfn id=storage-mutex>storage
-  mutex</dfn>. This mutex is used to control access to shared state
-  like cookies. At any one point, the <a href=#storage-mutex>storage mutex</a> is
-  either free, or owned by a particular <a href=#event-loop>event loop</a> or
-  instance of the <a href=#fetch title=fetch>fetching</a> algorithm.</p>
+  <p>A user agent may have one <dfn id=storage-mutex>storage mutex</dfn>. This mutex is
+  used to control access to shared state like cookies. At any one
+  point, the <a href=#storage-mutex>storage mutex</a> is either free, or owned by a
+  particular <a href=#event-loop>event loop</a> or instance of the <a href=#fetch title=fetch>fetching</a> algorithm.</p>
+
+  <p>If a user agent does not implement a <a href=#storage-mutex>storage mutex</a>,
+  it is exempt from implementing the requirements that require it to
+  acquire or release it.</p>
+
+  <p class=note>User agent implementors have to make a choice
+  between two evils. On the one hand, not implementing the storage
+  mutex means that there is a risk of data corruption: a site could,
+  for instance, try to read a cookie, increment its value, then write
+  it back out, using the new value of the cookie as a unique
+  identifier for the session; if the site does this twice in two
+  different browser windows at the same time, it might end up using
+  the same "unique" identifier for both sessions, with potentially
+  disastrous effects. On the other hand, implementing the storage
+  mutex has potentially serious performance implications: whenever a
+  site uses Web Storage or cookies, all other sites that try to use
+  Web Storage or cookies are blocked until the first site finishes.</p>
 
   <p>Whenever a <a href=#concept-script title=concept-script>script</a> calls into
   a <a href=#plugin>plugin</a>, and whenever a <a href=#plugin>plugin</a> calls into

diff --git a/source b/source
@@ -77351,11 +77351,28 @@ interface <dfn>NavigatorOnLine</dfn> {
   title="concept-script">script</span> then the document is the
   <span>script's document</span>.</p>
 
-  <p>A user agent is required to have one <dfn>storage
-  mutex</dfn>. This mutex is used to control access to shared state
-  like cookies. At any one point, the <span>storage mutex</span> is
-  either free, or owned by a particular <span>event loop</span> or
-  instance of the <span title="fetch">fetching</span> algorithm.</p>
+  <p>A user agent may have one <dfn>storage mutex</dfn>. This mutex is
+  used to control access to shared state like cookies. At any one
+  point, the <span>storage mutex</span> is either free, or owned by a
+  particular <span>event loop</span> or instance of the <span
+  title="fetch">fetching</span> algorithm.</p>
+
+  <p>If a user agent does not implement a <span>storage mutex</span>,
+  it is exempt from implementing the requirements that require it to
+  acquire or release it.</p>
+
+  <p class="note">User agent implementors have to make a choice
+  between two evils. On the one hand, not implementing the storage
+  mutex means that there is a risk of data corruption: a site could,
+  for instance, try to read a cookie, increment its value, then write
+  it back out, using the new value of the cookie as a unique
+  identifier for the session; if the site does this twice in two
+  different browser windows at the same time, it might end up using
+  the same "unique" identifier for both sessions, with potentially
+  disastrous effects. On the other hand, implementing the storage
+  mutex has potentially serious performance implications: whenever a
+  site uses Web Storage or cookies, all other sites that try to use
+  Web Storage or cookies are blocked until the first site finishes.</p>
 
   <p>Whenever a <span title="concept-script">script</span> calls into
   a <span>plugin</span>, and whenever a <span>plugin</span> calls into