search

zooniverse / operationwardiary

Diaries of the First World War
operationwardiary.org
1 stars 1 forks source link

Randomise subject and talk URLs if possible #7

Closed eatyourgreens closed 10 years ago

eatyourgreens commented 10 years ago

Luke can browse through the diary pages by incrementing the zooniverse_id in the Talk URL, or by incrementing the subject id in an image URL. Is it possible to randomise the identifiers and still classify the pages in sequence? We agreed with the National Archives that browsing through the diary pages by editing URLs wouldn't be possible.

ttfnrob commented 10 years ago

This means assigning random Zooniverse IDs - or at least seemingly random ones - to the subjects Talk Can this be done @parrish ?

parrish commented 10 years ago

Zooniverse id’s are auto incrementing. A non-zooniverse id based system could be implemented, but a fair amount of code depends on a standard zooniverse id.

Even if we create the subjects out of order to break exact sequencing, users are going to reorganize them in collections/discussions.

Also,

We’d have to implement a system that physically moved many thousands of files every five minutes. Given S3 request rates, this would require at least 3-4 EC2 instances. We’d also have to implement a hidden service that resolved those locations.

Between these two items, you’re talking about at least 2-3 weeks of development time and a lot of money to support it.

Now, if we used pre-signed S3 keys, we could kind of fake a changing url. Nothing is going to stop people from working around that though. Also you’re talking about adding a second or two to every image served in any response. If you’re viewing a collection with even 10 images in it, the pre-signed key will be expired before it even gets to the browser.

Honestly, if they aren't satisfied with a shuffled sequence on zooniverse ids, we have to disable Talk.

ttfnrob commented 10 years ago

Okay - understood. A semi-shuffled order will be just fine. It's just a low-level way of preventing people downloading the whole collection sequentially. You're right that they might just organise them anyway, but that's been discussed and is an acceptable risk.

parrish commented 10 years ago

Awesome. They'll be built in a randomized order for production.

ttfnrob commented 10 years ago

Thank you - it's appreciated.

On 6 January 2014 21:48, Michael Parrish notifications@github.com wrote:

Closed #7 https://github.com/zooniverse/Diaries/issues/7.

— Reply to this email directly or view it on GitHubhttps://github.com/zooniverse/Diaries/issues/7 .

eatyourgreens commented 10 years ago

Cool! Sorry, just catching up on this now. Quick question for @parrish: this won't break classifying the pages sequentially will it (because that works on subject ID)?

ttfnrob commented 10 years ago

I'm guessing that it will

On Tuesday, 7 January 2014, eatyourgreens wrote:

Cool! Sorry, just catching up on this now. Quick question for @parrishhttps://github.com/parrish: this won't break classifying the pages sequentially will it (because that works on subject ID)?

— Reply to this email directly or view it on GitHubhttps://github.com/zooniverse/Diaries/issues/7#issuecomment-31726962 .

Rob

mobile: +44(0)7929 508961 email: ttfnrob@gmail.com twitter: @orbitingfrog

ttfnrob commented 10 years ago

Can we just have two IDs? One for the classification interface and one for the rest? Jim - can you use another, existing key to order the diaries?

On Tuesday, 7 January 2014, Robert Simpson wrote:

I'm guessing that it will

On Tuesday, 7 January 2014, eatyourgreens wrote:

Cool! Sorry, just catching up on this now. Quick question for @parrishhttps://github.com/parrish: this won't break classifying the pages sequentially will it (because that works on subject ID)?

— Reply to this email directly or view it on GitHubhttps://github.com/zooniverse/Diaries/issues/7#issuecomment-31726962 .

Rob

mobile: +44(0)7929 508961 email: ttfnrob@gmail.com twitter: @orbitingfrog

eatyourgreens commented 10 years ago

At the moment the Zooniverse API is sending back the subjects in order eg. https://dev.zooniverse.org/projects/war_diary/groups/5241bcf43ae74068250005c7/subjects?limit=5

I don't think we can do the page ordering on the client side but everything should be fine if we're using subject ID to decide which subjects to send back for classification.