16 ms·

In reading through the blog, there's an article on the technical challenges of resizing the 135 million + images as you reorganized the site. Perhaps this is a

by bmelton 16y ago

In reading through the blog, there's an article on the technical challenges of resizing the 135 million + images as you reorganized the site. Perhaps this is a dumb question, but if you've already got a resize script that resizes images into 6 different sizes when the seller uploads it. What was the reason you didn't just re-point that script at the filesystem instead of an uploaded file object? It may have taken slightly longer to process, but my hunch is that it may have compensated for man-hours you spent in building something new. What was the limiting factor there?

code_duck 16y ago
Excellent question. Check out the comments. A few members of the community have questioned why they chose this elaborate process, which actually ended up with the new images actually having noticeably lower quality. This was noticed by the artists themselves, too, as Etsy's customers labor for a long time to get the highest quality photos possible. Etsy's response when this was brought up in the member forums was that they 'learned from this' and would take note for the future... a more satisfying response would be "we acknowledge that the images are inadequate and we'll re-do them". Why Etsy thought they had to seek new image algorithms rather than just using the ones they already used is the sort of question they very rarely answer. My analysis is that it is related to the fact that they did the processing on an Amazon server. To save bandwidth, transfer time and processing time, they sent over the medium sized images rather than the original, full size image. Of course, that doesn't explain why they decided to do it in 'the cloud' at all rather than, as you say, just operate directly on the file system, in house. They claim to have gone through an elaborate comparison process, apparently with the visual quality comparison being done by engineers - not people specializing in the non-technical side of photography or graphics. They seem to have entirely missed the quality issue, though it was apparent enough to customers that at least a hundred of them brought it up in the forums once the switch to the new image sizes was made.

by bmelton 16y ago

code_duck 16y ago

Excellent question. Check out the comments. A few members of the community have questioned why they chose this elaborate process, which actually ended up with the new images actually having noticeably lower quality. This was noticed by the artists themselves, too, as Etsy's customers labor for a long time to get the highest quality photos possible. Etsy's response when this was brought up in the member forums was that they 'learned from this' and would take note for the future... a more satisfying response would be "we acknowledge that the images are inadequate and we'll re-do them". Why Etsy thought they had to seek new image algorithms rather than just using the ones they already used is the sort of question they very rarely answer. My analysis is that it is related to the fact that they did the processing on an Amazon server. To save bandwidth, transfer time and processing time, they sent over the medium sized images rather than the original, full size image. Of course, that doesn't explain why they decided to do it in 'the cloud' at all rather than, as you say, just operate directly on the file system, in house. They claim to have gone through an elaborate comparison process, apparently with the visual quality comparison being done by engineers - not people specializing in the non-technical side of photography or graphics. They seem to have entirely missed the quality issue, though it was apparent enough to customers that at least a hundred of them brought it up in the forums once the switch to the new image sizes was made.