Upgrade StorageBetter communication about sync; faster infrastructure
Zotero is absurdly awesome and I cheerlead for it wherever I go. But sync is really slow, looks even slower, and the client doesn't do a good job of communicating where it is in the sync process or whether or not sync will ever actually work. It's scaring away my coworkers. "It doesn't work," they're saying. "It just sits there spinning forever."
I think there are a few interrelated causes for this negative perception:
a) People expect sync time, even at peak usage, to be measured in seconds, not minutes or hours. We've been trained by other software to see this as reasonable: dropbox, for example, handles similar data volumes much, much more quickly.
b) Zotero doesn't do anything to manage user expectations about sync speed. There's no time estimate, no queue position field, no data transfer bars. When a browser page sits spinning forever, it usually means the download will fail and we should give up and start over again. What's a user supposed to think about an arrow spinning into infinity without any real indication that it's working as expected?
I think that providing some information -- any information -- about status that gives the perception that Zotero is constantly phoning home and waiting for an update to the queue position would do a lot to help manage expectations and improve perceptions of the sync feature.
But the sync feature really is slow. I know servers and bandwidth are expensive, but this is almost unusable. (Yes, only during peak hours, but that's exactly when *I'm* trying to work, too.) I'm hoping that Zitation will be a way for us to get the performance we need without expecting your infrastructure to improve, but I get the sense that's a long way down the pipe as a mainstream option. If there were some kind of premium sync service that kept wait times low during peak hours, I'd sure as hell pay for it. If you can't afford to maintain a free service that provides acceptable performance, maybe you just shouldn't run a free service... it really damages the product.
Thanks everyone for all your hard work on this, and thanks for listening.
tim
I think there are a few interrelated causes for this negative perception:
a) People expect sync time, even at peak usage, to be measured in seconds, not minutes or hours. We've been trained by other software to see this as reasonable: dropbox, for example, handles similar data volumes much, much more quickly.
b) Zotero doesn't do anything to manage user expectations about sync speed. There's no time estimate, no queue position field, no data transfer bars. When a browser page sits spinning forever, it usually means the download will fail and we should give up and start over again. What's a user supposed to think about an arrow spinning into infinity without any real indication that it's working as expected?
I think that providing some information -- any information -- about status that gives the perception that Zotero is constantly phoning home and waiting for an update to the queue position would do a lot to help manage expectations and improve perceptions of the sync feature.
But the sync feature really is slow. I know servers and bandwidth are expensive, but this is almost unusable. (Yes, only during peak hours, but that's exactly when *I'm* trying to work, too.) I'm hoping that Zitation will be a way for us to get the performance we need without expecting your infrastructure to improve, but I get the sense that's a long way down the pipe as a mainstream option. If there were some kind of premium sync service that kept wait times low during peak hours, I'd sure as hell pay for it. If you can't afford to maintain a free service that provides acceptable performance, maybe you just shouldn't run a free service... it really damages the product.
Thanks everyone for all your hard work on this, and thanks for listening.
tim
-
bdarcus+1
-
dstillmanWe do too, and we're currently overhauling the sync architecture with this as a goal.
-
mlavalleeedited October 25, 2010Sorry to revive and old thread but : I think a queue position field would do wonders to user confidence. It does not need to be accurate, it just need to show that something is happening, that there is a progression, and that you should stop worrying and be patient. User tdsmith is right ; the common reflex of the web-user when faced with a seemingly neverending loading process is to assume something is wrong with the server.