Due to the space shuttle orbiter’s high weight (potentially up to 110 tonnes [121 short tons] in certain abort scenarios, although nominally around 85 tonnes [94 short tons]) and abysmal subsonic aerodynamics, it landed at very high speeds (nominally 195 to 205 KIAS1, corresponding to as much as 215 KGS with a ten-knot tailwind allowance2). As a result, a gear-up landing would have posed a considerable risk of voiding the warranties on the vehicle’s occupants:
Bailout is preferred over a gear-up landing attempt. [Space shuttle FCOM, page 6.9-2 (page 900 of the PDF of the manual); bolding in original.]
Despite the danger posed by a gear-up landing, the procedures for a normal shuttle landing did not call for extending the landing gear until on short final, approximately 300 feet AGL and a mere 20 or so seconds before ground impact; this would leave the crew with no good options in the event of a landing-gear-extension failure, forcing them to attempt a gear-up landing which would most likely destroy the orbiter and result in serious or fatal injuries to at least some of its crew (like we saw when we were trying to ditch the orbiter in the sea).
The safe deployment envelope for the shuttle’s landing gear opened much earlier than this; the gear were rated for safe deployment at up to 312 KIAS, and the shuttle, during a nominal approach, was already below this airspeed by the time it descended through 15 kft AMSL,3 and did not exceed it at any point thereafter. If the gear were commanded down at 15 kft instead of 300 feet, this would leave time, in the event of a deployment failure, to turn towards a sparsely-populated area, pull out to 185-195 KIAS (the recommended bailout airspeed), engage the autopilot, and (for at least some of the crew) abandon ship (rather than pushing over into the normal 300-KIAS dive down the outer MLS glideslope in preparation for landing).
Deploying the gear at 15 kft, rather than when 300 feet above the rabbit, would likely have required the outer glideslope to be steepened slightly to allow the orbiter to maintain the 300-knot target speed with the added drag of deployed landing gear,4 but this would have been an easy and simple adjustment for the MLS providing the glideslopes (unlike with the more common ILS, where such an adjustment would require complicated reconfiguration of the glideslope antennae, possibly including physically moving the antenna array).
So why was the shuttle’s landing gear held in until it was far too late for any sort of meaningful remedial action in the event that the gear failed to deploy when commanded to do so?
(Not a dupe of this question; theirs asks whether, mine asks why.)
1: The shuttle FCOM uses KEAS (Knots Equivalent Airspeed) rather than KIAS (Knots Indicated Airspeed), but the two should normally be equal (assuming that one’s flight instrumentation is properly maintained and calibrated and that one’s pitot-static system is functioning properly).
2: The maximum allowable groundspeed at main gear touchdown was 225 knots, beyond which the orbiter’s tyres could not be guaranteed to remain intact.
3: It’s quite possible that the shuttle orbiter never reached or exceeded 312 KIAS at any point during a nominal reentry, approach, and landing (and not all that surprising if so, given that this was only 21 knots below the shuttle’s VMO of 333 KIAS, a margin easily wiped out by moderate horizontal windshear); however, I can’t confirm this above 15 kft, as this is the first entry in the FCOM’s nominal-reentry-approach-and-landing timeline for which an airspeed value (rather than a mach number) is listed.
4: The portions of entry and approach prior to MLS acquisition (S-turning, air-data incorporation, transonic flight, transition to manual control, and most of the nominal-energy procedure turn5) would not be affected (except for being moved slightly closer to the runway to compensate for the steeper outer glideslope), as these took place before the vehicle descended through 15 kft; the inner glideslope would likewise have been unaffected, as even a 300-foot gear deployment would still see the orbiter flying down all but the very beginning of the inner glideslope with gear locked down (unless, of course, it failed to deploy). The minimum-energy procedure turn, performed closer to the runway than the nominal procedure turn and at a lower altitude, would likely have to be moved closer to the runway and flown at a higher descent rate to compensate for the higher drag from flying the procedure turn with gear deployed, but, again, ‘twould be but a simple tweak for the MLS.
5: Referred to by NASA as the “Heading Alignment Cone” (HAC), which is a fancy way of saying “procedure turn”.