<html xmlns:v="urn:schemas-microsoft-com:vml" xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:w="urn:schemas-microsoft-com:office:word" xmlns:m="http://schemas.microsoft.com/office/2004/12/omml" xmlns="http://www.w3.org/TR/REC-html40">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=us-ascii">
<meta name="Generator" content="Microsoft Word 15 (filtered medium)">
<base href="x-msg://6599/"><style><!--
/* Font Definitions */
@font-face
{font-family:"Cambria Math";
panose-1:2 4 5 3 5 4 6 3 2 4;}
@font-face
{font-family:Calibri;
panose-1:2 15 5 2 2 2 4 3 2 4;}
@font-face
{font-family:Verdana;
panose-1:2 11 6 4 3 5 4 4 2 4;}
@font-face
{font-family:Consolas;
panose-1:2 11 6 9 2 2 4 3 2 4;}
/* Style Definitions */
p.MsoNormal, li.MsoNormal, div.MsoNormal
{margin:0cm;
margin-bottom:.0001pt;
font-size:12.0pt;
font-family:"Times New Roman","serif";}
a:link, span.MsoHyperlink
{mso-style-priority:99;
color:blue;
text-decoration:underline;}
a:visited, span.MsoHyperlinkFollowed
{mso-style-priority:99;
color:purple;
text-decoration:underline;}
span.apple-style-span
{mso-style-name:apple-style-span;}
span.apple-converted-space
{mso-style-name:apple-converted-space;}
span.EmailStyle19
{mso-style-type:personal-reply;
font-family:"Verdana","sans-serif";
color:#1F497D;}
.MsoChpDefault
{mso-style-type:export-only;
font-size:10.0pt;}
@page WordSection1
{size:612.0pt 792.0pt;
margin:72.0pt 72.0pt 72.0pt 72.0pt;}
div.WordSection1
{page:WordSection1;}
--></style><!--[if gte mso 9]><xml>
<o:shapedefaults v:ext="edit" spidmax="1026" />
</xml><![endif]--><!--[if gte mso 9]><xml>
<o:shapelayout v:ext="edit">
<o:idmap v:ext="edit" data="1" />
</o:shapelayout></xml><![endif]-->
</head>
<body lang="EN-GB" link="blue" vlink="purple">
<div class="WordSection1">
<p class="MsoNormal" style="margin-left:36.0pt">What it would mean is that, if we do decide to generalize roles further, users would have one release that allows noticeably fewer uses of GND than the one before or the one after it. On the other hand, if we're
confident that we're not going to tinker with roles more, then I see much stronger reasons for releasing with 7.8.<o:p></o:p></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Verdana","sans-serif";color:#1F497D;mso-fareast-language:EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Verdana","sans-serif";color:#1F497D;mso-fareast-language:EN-US">I don’t think we know for sure. But the way to find out is to release it. If we don’t have a compiler that supports roles, no one
will explore or use them. We do this all the time, and yes, Haskell is a bit less smoothly-upgradable as a result. But I think it’s worth it, otherwise we’d be stuck in treacle.<o:p></o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Verdana","sans-serif";color:#1F497D;mso-fareast-language:EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Verdana","sans-serif";color:#1F497D;mso-fareast-language:EN-US">Remember this is fixing a current unsoundness that can lead to un-caught segmentation faults in user code. That is a Jolly Good
thing to fix, it not just a snazzy feature that two people want.<o:p></o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Verdana","sans-serif";color:#1F497D;mso-fareast-language:EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Verdana","sans-serif";color:#1F497D;mso-fareast-language:EN-US">Richard, how are you fixed at the moment? Are you able to implement the new GND check soon? Or not?<o:p></o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Verdana","sans-serif";color:#1F497D;mso-fareast-language:EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Verdana","sans-serif";color:#1F497D;mso-fareast-language:EN-US">Simon<o:p></o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Verdana","sans-serif";color:#1F497D;mso-fareast-language:EN-US"><o:p> </o:p></span></p>
<div style="border:none;border-left:solid blue 1.5pt;padding:0cm 0cm 0cm 4.0pt">
<div>
<div style="border:none;border-top:solid #E1E1E1 1.0pt;padding:3.0pt 0cm 0cm 0cm">
<p class="MsoNormal"><b><span lang="EN-US" style="font-size:11.0pt;font-family:"Calibri","sans-serif"">From:</span></b><span lang="EN-US" style="font-size:11.0pt;font-family:"Calibri","sans-serif""> Richard Eisenberg [mailto:eir@cis.upenn.edu]
<br>
<b>Sent:</b> 16 October 2013 18:29<br>
<b>To:</b> Simon Peyton-Jones<br>
<b>Cc:</b> Edward Kmett; ghc-devs<br>
<b>Subject:</b> Re: More GND + role inference woes<o:p></o:p></span></p>
</div>
</div>
<p class="MsoNormal"><o:p> </o:p></p>
<div>
<p class="MsoNormal">Please see my responses below. (Note: this was written before Edward's most recent post, which will be addressed shortly.)<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal">Executive summary: Edward's problem is more general than the one that Simon solves, but what Edward wants is actually not type safe.<o:p></o:p></p>
</div>
<p class="MsoNormal"><o:p> </o:p></p>
<div>
<div>
<p class="MsoNormal">On Oct 16, 2013, at 9:28 AM, Simon Peyton-Jones wrote:<o:p></o:p></p>
</div>
<p class="MsoNormal"><br>
<br>
<o:p></o:p></p>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">I think I know what to do about the GND + roles question.</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><b><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">First, the problem</span></b><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">. Here’s an example:<br>
</span><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D"> class Trans t a where</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D"> foo :: t a -> t a</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D"> newtype N x = MkN x deriving( Trans Maybe )</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">As things stand,<span class="apple-converted-space"> </span></span><b><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D">Trans</span></b><span class="apple-converted-space"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span></span><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">gets
roles (representational, nominal). The second parameter ‘a’ gets a nominal role because we don’t know what ‘t’ will get instantiated too.</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">As a result the attempted GND is rejected because of the new role stuff. But with 7.6.3 we’ll get an instance</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D"> instance Trans Maybe a => Trans Maybe (N a)</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">from the GND mechanism.</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Do I have this right?</span><o:p></o:p></p>
</div>
</div>
</blockquote>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">This is a more specific instance of the problem. Yes, the case with `Trans` does appear, but also cases like this:<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<p class="MsoNormal"><span style="font-family:Consolas">class Functor f => Core f where</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-family:Consolas"> core :: ((forall g x. Functor g => (x -> g x) -> f x -> g (f x)) -> a) -> f a</span><o:p></o:p></p>
</div>
</blockquote>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
</div>
<div>
<p class="MsoNormal">That's a nasty-looking type, but note the `g (f x)` that appears. Here, `f` is used as a parameter to a type variable that *is not a class variable*. The problem Simon describes is only the case where the applied type variable is a class
variable, and thus is known at `deriving` time.<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">To handle things like Core, we would need to have more proper role abstraction, where we can assert that `g` has a parameter with a representational role in order to infer `f` as a representational parameter for Core. In Edward's `linear`
package, the Linear.Affine module has this definition:<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<p class="MsoNormal"><span style="font-family:Consolas">newtype Point f a = P (f a)</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-family:Consolas"> deriving ( Eq, Ord, Show, Read, Monad, Functor, Applicative, Foldable</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-family:Consolas"> , Traversable, Apply, Bind, Additive, Metric, Core, R1, R2, R3, R4</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-family:Consolas"> , Fractional , Num, Ix, Storable, Epsilon</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-family:Consolas"> )</span><o:p></o:p></p>
</div>
</blockquote>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">Of these derivings, the following currently fail: Traversable, Bind, Core, R1, R2, R3, and R4. It is straightforward to fix Traversable with DeriveTraversable (and a standalone instance). Bind will fixed by the change in the GND check.
But, Core and the Rs won't be and would have to be written by hand.<o:p></o:p></p>
</div>
</div>
<p class="MsoNormal"><br>
<br>
<o:p></o:p></p>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><b><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Second, there is a Good Reason for the problem</span></b><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">. Suppose I said</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> newtype N x = MkN x deriving( Trans D )</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">where D was a data family. Then the foo method for (D x) might seg-fault if applied to a (D (N x)) value. So the current GND is treading on very thin ice,
and it is ice that the author (Edward) does not control. His clients control it.</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Do I have this right?</span><o:p></o:p></p>
</div>
</div>
</blockquote>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">Yes.<o:p></o:p></p>
</div>
<p class="MsoNormal"><br>
<br>
<o:p></o:p></p>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><b><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Third, there is an easy solution</span></b><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">. As Richard observed in a separate
thread about Coercible, GND does not attempt to build a coercion to witness</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D"> Trans Maybe x ~R Trans Maybe (N x)</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Rather, it builds a dictionary for (Trans Maybe (N x)) thus</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D"> dTransMaybeN :: Trans Maybe x -> Trans Maybe (N x)</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D"> dTransMaybeN d = MkTrans (sel_foo d |> co)</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">where the (sel_foo d) selects the foo method from d. That is, we actually cast the methods, not the dictionary as a whole. So what we need is a coercion
between</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D"> Maybe x ~R Maybe (N x)</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Can we get that? Yes of course! Maybe has a representational role.</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><b><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Bottom line: we should accept or reject a GND attempt not on the basis of the role of the class, but rather on whether the method types are coercible. <span class="apple-converted-space"> </span></span></b><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Indeed,
this is just what we agreed some days ago</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"><a href="http://www.haskell.org/pipermail/glasgow-haskell-users/2013-October/024386.html">http://www.haskell.org/pipermail/glasgow-haskell-users/2013-October/024386.html</a></span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">So I think this solves Edward’s problem below.</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Moreover, I think this solves the other failures in<span class="apple-converted-space"> </span><a href="http://www.haskell.org/pipermail/ghc-devs/2013-October/002961.html">http://www.haskell.org/pipermail/ghc-devs/2013-October/002961.html</a>.
Here is one example, in that email. smallcheck has this:</span><o:p></o:p></p>
</div>
<div style="margin-left:36.0pt">
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D">newtype Series m a = Series (ReaderT Depth (LogicT m) a)</span><o:p></o:p></p>
</div>
<div style="margin-left:36.0pt">
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D"> deriving</span><o:p></o:p></p>
</div>
<div style="margin-left:36.0pt">
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D"> ( …, MonadLogic)</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">where logict defines MonadLogic thus:</span><o:p></o:p></p>
</div>
<div style="margin-left:36.0pt">
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div style="margin-left:36.0pt">
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D">class (MonadPlus m) => MonadLogic m where</span><o:p></o:p></p>
</div>
<div style="margin-left:36.0pt">
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D"> msplit :: m a -> m (Maybe (a, m a))</span><o:p></o:p></p>
</div>
<div style="margin-left:36.0pt">
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">So the “bottom line” check above will attempt to find a cocercion betwem msplit’s type with m=Series m, and with m=ReaderT Depth (LogitT m). Right?</span><o:p></o:p></p>
</div>
</div>
</blockquote>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">Yes.<o:p></o:p></p>
</div>
<p class="MsoNormal"><br>
<br>
<o:p></o:p></p>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> So on the left of msplit’s arrow, we’ll ask can we prove</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1F497D"> Series m a ~R ReaderT Depth (LogicT m) a</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Can we show that? Yes, of course… that is the very newtype coercion for Series.</span><o:p></o:p></p>
</div>
</div>
</blockquote>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">Well, it's the right-hand side of the arrow that's more troublesome, but that works out in this case, too.<o:p></o:p></p>
</div>
<p class="MsoNormal"><br>
<br>
<o:p></o:p></p>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">In short, I think we are fine, once Richard has implemented the new GND test.</span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Am I wrong?</span><o:p></o:p></p>
</div>
</div>
</blockquote>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">It depends on what "fine" is.<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">The new GND check will allow more deriving, but not all that Edward wants. BUT, the ones that (the new) GND doesn't allow aren't actually safe! In the Point example, using GND on Core and the Rs might actually lead to a type error, depending
on the instantiation of local type variables within the methods in those classes. For example, if the `g` variable in Core's `core` method is instantiated with a type with a nominal parameter, there could be trouble.<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">So, I disagree with Edward's original claim at the top of this thread that roles cause havoc here: they don't cause havoc, they catch latent type errors!<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">On the other hand, I agree that with more machinery (that is, a constraint on `g` about its parameter's role) we could potentially allow all the derived instances that Edward wants.<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">What I'm worried about is the user experience like the following:<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal">- 7.8 comes out: A bunch of deriving clauses have to be expanded to hand-written instances. Annoying, and with runtime cost.<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal">- 7.10 comes out: A bunch of those expanded instances now can be rolled back into derived ones, with some extra, not-backward-compatible annotations (that is, the constraints on variables like `g`). Annoying, and a real maintenance headache.<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">What it would mean is that, if we do decide to generalize roles further, users would have one release that allows noticeably fewer uses of GND than the one before or the one after it. On the other hand, if we're confident that we're not
going to tinker with roles more, then I see much stronger reasons for releasing with 7.8.<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">Richard<o:p></o:p></p>
</div>
</div>
</div>
</div>
</body>
</html>