I've noticed that a previously answered question has recently become open again: where does a typical user place their threshold for a download that's too big?
Companies that distribute client software via a web download used to agonize over how to reduce their download size in order to encourage adoption. Browser developers in particular used to crow about how their download was smaller than their competitors. This was a big deal in the days of dialup. If I recall, a rule of thumb in the mid 90s held that each 1MB of download would take 10 minutes over a typical dialup connection of that period. Worse, dialup connections could easily fail, forcing the user to sit through the long download again. A user who had to devote an hour of their life to babysit your 6MB download was quite cautious about clicking the "Download Now" button. The user was also concerned about how much disk space your product would ultimately eat up was everything was installed.
The relevant factors have changed significantly: many users have broadband connections, download managers exist to cope with flaky connections, and hard drive space is easy enough to come by. Many companies now seem to pay scant attention to their download size, so I can only assume many users don't care either. Download Adobe's Reader product (a business requirement these days) and you'll see a lightning quick 500K download... of Adobe's Download Manager, which then brings down another 80MB or so of software—including some other Adobe products slipped in for good measure.
One reason this question is interesting to client software designers is that there are some pretty interesting client runtimes coming down the pike that open up some great UI possibilities at the expense of download and install size. The forthcoming .NET Framework version 2.0 package is 22MB. The Windows Presentation Foundation ("Avalon") or Windows Communcation Foundation ("Indigo") add more on top of that. If you're building a product that itself already depends on broadband use, does the download size of your client even matter at all? If anyone has actual, recent data (not anecdotal experiences) on how todays user's react to download size, I'm interested in hearing it.
Solution: Use Thinstall
Posted by: RichB | September 06, 2005 at 04:15 AM
I'd say it depends. 10MB for Google Earth seems surprisingly small, whereas Quicktime 7 (even without iTunes) is a mind-blowing 35MB, and as you mentioned, Acrobat is far larger than you'd imagine. I'd say the file size should be in line with what people expect the app to do. A "simple" app that views PDF files should be no more than a couple megabytes at most. See the tiny Flash Player for a good example. On the other hand, a fully-featured piece of software like Picasa might be able to get away with much more.
Posted by: Ben | September 10, 2005 at 06:56 PM
I work for a large consumer software company that has several million copies of our client software installed. We still find that there is a direct relationship between download size and aborted install; you see significant losses above 5 MB.
I asked a friend on the CLR team what MS's best development solution was for an app that needed to have a small footprint and widespread compatability while maintaining ease of development. After thinking about it for a minute he said, "We don't have one."
Posted by: Dan S | November 05, 2005 at 01:52 PM