CAPE CANAVERAL, SEPT. 6 -- While frustrated NASA engineers wrestled with the hydrogen fuel leak that prevented today's scheduled launch of Columbia, a different and potentially major leak emerged to threaten the high-priority October launch of the shuttle Discovery on the Ulysses mission to study the sun.

"We need a win, somewhere," said unhappy launch director Robert Seick.

A leak of ammonia in a part of Discovery's cooling system "may eat into the front end of the {launch} window for Ulysses," National Aeronautics and Space Administration Administrator Richard H. Truly told reporters at a briefing here. He was referring to the tight 18-day window of opportunity for starting the joint U.S.-European mission, which he said is NASA's top priority for the rest of the year.

The Ulysses spacecraft must be launched by Oct. 23 or miss an alignment of the planets that will not recur for 13 months.

So crucial was the Ulysses mission, aboard Discovery, that officials said they would not threaten it by allowing much delay in the launch of Columbia. But now that Discovery's launch is delayed, there may be time to repair Columbia's hydrogen leak and squeeze in a fourth attempt to launch its astronomy mission, called Astro-1, early the week of Sept. 17, he said.

However, one engineer said, if the worst-case situation arose on the attempt to repair Discovery's ammonia leak -- if the entire fitting must be replaced -- the work could require three extra weeks, on top of routine launch preparations, and consume virtually the entire launch window. "It could get dicey," he said and Columbia might have to wait until after Discovery goes.

The hope is that the leak is in a place that is easier, and quicker, to fix.

The part that may have to be replaced, called an ammonia boiler, is required to cool the orbiter cabin during an emergency landing.

NASA officials have been hard-pressed to restore momentum and put a good face on their four-month run of bad news and bad breaks, including a major flaw in the Hubble Space Telescope; grounding of the shuttle fleet for the summer by a series of hydrogen leaks; unexplained communications lapses with the Magellan probe orbiting Venus, and a spate of technical and other problems in development of the proposed space station. These developments have jeopardized NASA's budget request on Capitol Hill.

"This has been a troublesome and vexing summer for me and for NASA," Truly said, repeating his defense of the agency and emphasizing the difficulty of its assignment -- "to push the boundaries of human activity."

"But let me tell you what we're not going to do," he added, alluding to the loss of the Challenger in 1986: "Ask or allow the team to be pushed" into an unsafe launch.

Engineers were increasingly optimistic about chances of fixing Columbia's leak, they said, as data indicated its location -- a recirculation pump that uses some of the hydrogen to chill the main engine before it receives the full flow of liquid hydrogen at ignition. The part can be replaced relatively easily on the launch pad, officials said.

There is a good chance, "probably better than 50-50," that the leak is in the pump, and that replacing it will fix the problem, said NASA chief of space flight William Lenoir.

The culprit that has given NASA fits all summer is the tiniest molecule in the universe -- the hydrogen molecule which is made up of two atoms of hydrogen, smallest of all atoms.

To contain the tricky and dangerous hydrogen, shuttle engineers face problems of any plumbing system -- but with major complications. The connector pipe that carries fuel from the external tank into the orbiter main engines is large, 17 inches across.

In addition, the pipes are covered with insulating foam. When hydrogen leaks, it can migrate through the foam and emerge anywhere else, confounding efforts to find the original hole.

Moreover, the cold temperatures cause the complex tangle of shuttle plumbing to shrink from the surrounding seals.

Some leakage under these conditions is inevitable, engineers say, and some expressed surprise that in the shuttle's history there have not been bigger problems.

"Almost more troubling" than the leaks, Lenoir said, is the fact that the batteries of tests and inspections set up through the process to catch them have not caught them.

The leak in Columbia, and a similar one found in June in the shuttle Atlantis, was traced to a fitting known as the 17-inch "disconnect," which carries fuel from the giant external tank into the belly of the orbiter and its engine compartment.

The hydrogen molecule's tendency to migrate through the foam has fooled engineers twice.

First, they thought the first leak on Discovery was from one source, and had migrated into the engine compartment from a leak between the tank and the orbiter. It turns out there was a second leak in the engine compartment.

Second, they thought Atlantis's leak in June was not in the disconnect because it appeared to be coming from a different place, and to be a separate problem. They were wrong.

When Columbia's disconnect was removed and shipped to California for testing, engineers traced the leakage to small seals around shafts that open and close valves in the fitting.

Technicians found that up to 200 microscopic glass beads had gotten into the assembly, possibly scratching grooves in the seals that allowed the hydrogen to escape. The beads were found in another disconnect manufactured at the same time in the early 1980s.

Parker Hannifin Corp. made the disconnects. It was using the beads for unrelated projects in the same building where the shuttle parts are made, Lenoir said. But the experts have not figured out how the beads got into the shuttle plumbing.

Columbia's disconnect fitting was replaced with one from Endeavour, the new shuttle under construction, and it did not leak during Wednesday's fueling.

Engineers do not believe Columbia's latest leak, in the engine compartment, is related to the bead problem, officials said. The leak could have been created when the pump package was removed for testing following an earlier launch, Lenoir said.

If Columbia cannot be fixed in time to launch before Discovery, officials said, it will be next in line for lift-off following Ulysses.