Philosophy aside, I am getting very inconsistent results so far in having the DCS-930L send (upon motion capture) jpegs to my router-hosted USB flash memory, while trying to monitor what is going on via browser in the Mac.
The router+USB hookup has flawlessly handled GB+ sized transmissions over the web (folders of photos), so it seems that part of the link is working properly. The router is set to give the camera a static IP; the camera set to accept the router's DHCP service. My DCS-930L firmware is v1.00 - perhaps I should update? Camera setting selections for ftp are "Enable uploading ...", "motion detection", and "Date/Time Suffix", with "Creat Subfolder" hourly. The video is set to 15 fps currently, but results seem to be as bad no matter how I set it. I notice that the data rate readout on the mydlink mobile app seems not to track the fps setting.
I've been using the Stainless.app as an auxiliary browser, but looking at some camera page source, I see that there is coded-in reference to certain specific [MS-centric] browsers! This may explain why the browser often reports incorrect page names in the tabs, such as xxxx/wizard.htm in the tab while the address bar is set at xxxx/jview.htm for live video. Other strange things - I have observed the time readout in the "Status" panel incrementing, while unable to get video from ip/video.cgi in another tab; and at the same time a connected-device sensor (WakeOnLan.app) thinks the camera IP address is disconnected. A seemingly impossible contradiction.
I don't think I have the one-minute problem; perhaps the seemingly random dropouts via mydlink to the mobile device are a separate issue, since that is a stream via mydlink's server vs. the virtually local storage triggered by occasional action of the motion detector.
I am wondering why you believe that the camera can handle multiple requests - is this documented? There must be some limits imposed by the tiny processor in the camera.