Hi Edwin, Your last email set off a flurry of design discussion here in Mountain View. I think your point about async noun calls has hit on a serious problem. We came up with several possible ways to combat this, but in general I think that a better solution is to solve the underlying problem of how often nouns are being queried, rather than to give opt-outs to individual nouns. Since this is of interest mainly to people working on core ubiquity development, I'm going to start a thread on the ubiquity-core mailing list ([email protected]) to discuss possible strategies for making orders-of-magnitude cuts in the number of network calls being sent out. Of course, anyone who wants is welcome to come over to that list to participate in the discussion. It's just going to get very technical very fast since we'll have to go deep into the parser algorithms to figure this out. --Jono
On Jun 25, 4:54 pm, Edwin Khodabakchian <[email protected]> wrote: > Hi Jono, > > Thanks for your responsiveness. > > Regarding nouns, if this is done by design and is necessary, then > please provide us with a configuration knob on the noun to allow some > nouns not to participate into this and only be activated when the > command has been selected: all our nouns are async and even with the > latest xpi, there is a noun invocation on every character entry - I do > not see how you could otherwise make such a pattern scale. > > Regarding selection we have a save-for-later command with an alias > which can include a set of tags. If I type: > save-for-later test as qa, tag2 > the parser will select the option where the object is test as qa, > tag2. The second option is object:= test, alias := qa, tag2 If I > select that and continue typing to add more tags, it will go back to > automatically selecting the first option. > > -Edwin > > On Jun 25, 4:31 pm, Jono DiCarlo <[email protected]> wrote: > > > Hi Edwin, > > Thanks for your feedback. > > > > 1) feedback regarding API compatibility. Next time you are breaking > > > the API, I think that it would be nice to not automatically upgrade > > > the users from 0.2 to 0.5. > > > You are right to be concerned. We didn't break backwards > > compatibility lightly. We tried to get *everything* forseeable that > > would break backwards compatibility together and do it all in this one > > release, so we don't think we'll ever have to do another one that > > breaks it. I blogged more about this issue > > here:http://jonoscript.wordpress.com/2009/06/25/ubiquity-0-5-preview-relea... > > > > 2) feedback regarding behavior. In 0.5 I see a bizarre behavior where > > > nouns are invoked as the user starts typing (even before a command is > > > selected). > > > That's working as designed -- it is trying to analyze the input to see > > if it matches any nountypes; if it does it will produce verb > > suggestions based on the nouns. This is to help with the case where > > the user selects text and brings up ubiquity; the case where the user > > selects text and uses the context menu; cases where no verbs match the > > input; and for the benefit of languages that have their nouns before > > their verbs in normal sentence order. > > > > This is problematic from a performance perspective when > > > async nouns are defined - ends up generating a lot of unnecessary http > > > requests and callbacks. > > > About an hour ago, I pushed a new version of Ubiquity (0.5pre2) that > > includes a performance fix Brandon did > > inhttps://ubiquity.mozilla.com/hg/ubiquity-firefox/rev/582e0e7ca551. I > > haven't publicized it yet, but if you > > re-downloadhttp://ubiquity.mozilla.com/xpi/ubiquity-latest-beta.xpi now, you > > should get 0.5pre2. > > > You are right that the change in algorithm does do more calls than the > > old one used to, but it has a lot of other advantages, and with the > > last fix the performance hit has been minimized. > > > > 3) when the user selects a command which includes and object and an > > > alias, the parser seems to select by default the option where the > > > object encapsulates everything. If the user selects the second options > > > and continues typing, the parser does not seem to remember the user > > > choice and reselects the first option in the list. > > > I'm not sure what you mean by "a command which includes an object and > > an alias". Can you please give an example? (What you input, what > > gets suggested, what you think should be suggested instead.) > > > Thanks, > > --Jono, Mozilla Labs --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "ubiquity-firefox" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/ubiquity-firefox?hl=en -~----------~----~----~----~------~----~------~--~---
