[darcs-users] darcs UI: send = bundle file; send --mail = mail

Eric Kow eric.kow at gmail.com
Sat Aug 18 10:11:42 UTC 2012


On 15 Aug 2012, at 20:46, Gabriel Kerneis wrote:
> On Wed, Aug 15, 2012 at 01:18:28PM +0100, Eric Kow wrote:
>> As I currently understand the situation, the
>> default-configed-mailer-I-didn't-even-know-was-there *thinks* it succeeded.
>> So I suppose we need some way of gathering extrinsic evidence that the send
>> worked (and this in the context of trying to send a patch, not go through some
>> kind of darcs configuration process)
> 
> CC the default email address of the user doing darcs send, and print a message
> saying: "I've cc'ed you (user at example.com), if you don't receive a copy of the
> patch, consider doing blablabla...".

Well, I think this is helping us to grasp a bit at two of core aspects of the problem:

1. intrinsic delay between action and feedback
2. lack of failure-feedback != success-feedback  

What I mean is that when it comes to sending things by email, there is this inherent gap between when we send the mail, and our user gets confirmation that the mail was received.  CC'ing the user may reduce the gap, but only by so much, because we still rely on the user to remember to open their mail client, check that they receive the CC, etc.  The delay is intrinsic due to the asynchronous nature of email.  Short of Darcs itself taking the trouble to somehow check for a receive receipt and remind you the next time you use Darcs “hey, this mail doesn't seem like it was received in the last week…” I can't see what we can do about that intrinsic delay.

Moreover we have still have the core problem of absence of feedback.  Not getting feedback that message was not received is not the same as getting feedback that the message was received.  The latter is completely passive; the former requires cognitive effort (users have to go out of their way to notice the absence of feedback).  CC'ing the user does not really help very much IMHO because we still find ourselves having to rely on them to apply the cognitive effort to remember to check that they actually received the CC.  The more likely scenario (speaking as a highly dis-SQUIRREL!-ractible guy) is that the user forgets to check in the intervening time.

Well as Ben says, it's not foolproof, and I don't mean to imply that we should reject any solution that isn't foolproof. Solving the problem may involve incorporating the CC-to-self trick.  I just think that we're getting a bit closer to figuring out what the heart of the problem is, which will hopefully help us to get a better solution.

Man getting UI right is hard!
I hope I'm not being unnecessarily difficult about this! :-)
I'm sure we'll work something out over time.

-- 
Eric Kow <http://erickow.com>



More information about the darcs-users mailing list