Interesting. We can *almost* do this. You can set the permission on any +*read card on a Task card, but the permission for, say, Task1 isn't "Task1+*read"; it's "Task1+*self+*read". That's not yet an addressable Set. Conceptually this would be a rather simple mod, though; you'd just have to add another set pattern.
Once you had the Set in place, setting the read rule to "anyone signed in" would be easy. Setting it to _creator, however, would require its own customization, at least until we support owner permissions.
I was able to add {{+*self+*read}} to the task+*type+*structure and it worked as expected, but only the admin could see it. I wasn't sure which card to alter to give logged in users access to the task+*type+*self+*read (if that even is the correct spot...)
Owner permissions would be helpful because I wasn't able to get the *read+*right+*default card to accept {{_user}} as a pointer value, it worked sometimes, but wasn't consistent...
what you want to let them edit is all the cards that fit this pattern: [taskname]+*self+*read. But because of the current limitation of Sets, you can't attach rules to that and only that group of cards. You can address "the set of all cards ending in +*read", but that's too general. And you can say who can edit a single read rule, but that's too specific.
I'm super excited about pushing the Sets concept further, but I don't think we'll see a ton of change there before 2.0.
I attempted to make a +permissions card with a +*right+*structure that contained {{_left+*self+*read}} which allowed a user to set the task card when +permission was set in the task structure. However that didn't work. I guess I can't have it all :(
giving them permission to edit the +permission won't give them permission to edit the *read card. (Otherwise you could always gain edit permission by including a card, which would obviously be a gaping security hole). This is a good idea, and you could get it to work, but not without opening a lot more +*read cards to editing than you want. It becomes a community trust issue, I suppose. I generally like a lot of openness, but this would give casual users the opportunity to say that nobody can read the site. That idea pushes my boundaries a bit, not because of malice so much as the fact that it makes it just a bit too easy to make really ugly accidents.